Industrial Separation Processes: Fundamentals 9783110306729, 9783110306699

Table of contents :
1 Characteristics of separation processes
1.1 Significance of separations
1.2 Characteristics of separation processes
1.2.1 Categorization
1.2.2 Separating agents
1.2.3 Separation factors
1.3 Industrial separation methods
1.3.1 Exploitable properties
1.3.2 Important molecular separations
1.4 Inherent selectivities
1.4.1 Equilibrium based processes
1.4.2 Rate controlled processes
2 Evaporation and distillation
2.1 Separation by evaporation
2.1.1 Introduction
2.1.2 Vapor-liquid equilibria
2.2 Separation by single-stage partial evaporation
2.2.1 Differential distillation
2.2.2 Flash distillation
2.2.2.1 Specifying T and L/V
2.2.2.2 Specifying T and Ptot
2.3 Multistage distillation
2.3.1 Distillation cascades
2.3.2 Column distillation
2.3.3 Feasible distillation conditions
2.3.4 External column balances
2.4 McCabe-Thiele analysis
2.4.1 Internal balances
2.4.1.1 Rectifying section
2.4.1.2 Stripping section
2.4.1.3 Feed stage considerations
2.4.1.4 Feed line
2.4.2 Required number of equilibrium stages
2.4.2.1 Graphical determination of stages and location of feed stage
2.4.2.2 Limiting conditions
2.4.2.3 Fenske and underwood equations
2.4.2.4 Use of murphree efficiency
2.4.3 Energy requirements
2.5 Advanced distillation techniques
2.5.1 Batch distillation
2.5.2 Continuous separation of multiple product mixtures
2.5.3 Separation of azeotropes
3 Absorption and stripping
3.1 Introduction
3.2 The aim of absorption
3.3 General design approach
3.4 Absorption and stripping equilibria
3.4.1 Gaseous solute solubilities
3.4.2 Minimum absorbent flow
3.5 Absorber and stripper design
3.5.1 Operating lines for absorption
3.5.2 Stripping analysis
3.5.3 Analytical kremser solution
3.6 Industrial absorbers
3.6.1 Packed columns
3.6.2 Plate columns
3.6.3 Spray and bubble columns
3.6.4 Comparison of absorption columns
4 General design of gas/liquid contactors
4.1 Introduction
4.2 Modeling mass-transfer
4.3 Plate columns
4.3.1 Dimensioning a tray column
4.3.2 Height of a tray column
4.3.3 Diameter of a tray column
4.4 Packed columns
4.4.1 Random packing
4.4.2 Structured packing
4.4.3 Dimensioning a packed column
4.4.4 Height of a packed column
4.4.5 Minimum column diameter
4.4.6 Pressure drop
4.5 Criteria for column selection
5 Liquid-liquid extraction
5.1 Liquid-liquid extraction
5.1.1 Introduction
5.1.2 Liquid-liquid equilibria
5.1.3 Solvent selection
5.2 Extraction schemes
5.2.1 Single equilibrium stage
5.2.2 Cocurrent cascade
5.2.3 Crosscurrent
5.2.4 Countercurrent
5.3 Design of countercurrent extractions
5.3.1 Graphical McCabe-Thiele method for immiscible systems
5.3.2 Analytical kremser method
5.3.3 Graphical method for partial miscible systems
5.3.4 Efficiency of an ideal non-equilibrium mixer
5.4 Industrial liquid-liquid extractors
5.4.1 Mixer-settlers
5.4.2 Mechanically agitated columns
5.4.3 Unagitated and pulsed columns
5.4.4 Centrifugal extractors
5.4.5 Selection of an extractor
6 Adsorption and ion exchange
6.1 Introduction
6.2 Adsorption fundamentals
6.2.1 Industrial adsorbents
6.2.2 Equilibria
6.2.3 Kinetics
6.3 Fixed-Bed adsorption
6.3.1 Bed profiles and breakthrough curves
6.3.2 Equilibrium theory model
6.3.3 Modeling of mass transfer effects
6.4 Basic adsorption cycles
6.4.1 Temperature-swing
6.4.2 Pressure-swing
6.4.3 Inert and displacement purge cycles
6.5 Principles of ion exchange
6.5.1 Ion exchange resins
6.5.2 Equilibria and selectivity
6.6 Ion exchange processes
7 Drying of solids
7.1 Introduction
7.2 Humidity definitions of carrier gas
7.2.1 Definitions
7.2.2 System air-water: a special case
7.2.3 Psychrometric charts
7.3 Moisture in solids
7.3.1 Bound and unbound water
7.4 Drying mechanisms
7.4.1 Constant drying rate
7.4.2 Falling drying rate
7.4.3 Estimation of drying time
7.5 Classification of drying operations
7.5.1 Direct-Heat dryers
7.5.2 Contact dryers
7.5.3 Other drying methods
8 Crystallization and precipitation
8.1 Introduction
8.2 Crystal characteristics
8.2.1 Morphology
8.2.2 Crystal size distribution
8.3 Crystallization operating modes from solutions
8.3.1 Operating modes
8.3.2 Cooling crystallizers
8.3.3 Evaporating and vacuum crystallizers
8.3.4 Continuous crystallizers
8.3.5 Basic yield calculations
8.4 Crystallizer modeling and design
8.4.1 Supersaturation and metastability
8.4.2 Nucleation
8.4.3 Crystal growth
8.4.4 Population balance equations for the MSMPR crystallizer
8.5 Other crystallization techniques
8.5.1 Precipitation
8.5.2 Melt crystallization
9 Sedimentation and settling
9.1 Introduction
9.2 Gravity sedimentation
9.2.1 Sedimentation mechanisms
9.2.2 Dilute sedimentation
9.2.3 Hindered settling
9.2.4 Continuous sedimentation tank (gravity settling tank)
9.2.5 Gravity sedimentation equipment
9.3 Centrifugal sedimentation
9.3.1 Particle velocity in a centrifugal field
9.3.2 Sedimenting centrifuges
9.3.3 Bowl centrifuge separation capability
9.3.4 The sigma concept
9.3.5 Capacity of disc centrifuges
9.3.6 Hydrocyclones
9.4 Electrostatic precipitation
9.4.1 Principles
9.4.2 Equipment and collecting efficiency
10 Filtration
10.1 The filtration processfiltration process
10.2 Filtration fundamentals
10.2.1 Flow through packed beds
10.2.2 Cake filtration
10.2.3 Constant pressure and constant rate filtration
10.2.4 Compressible cakes
10.3 Filtration equipment
10.3.1 Continuous large-scale vacuum filters
10.3.2 Batch vacuum filters
10.3.3 Pressure filters
10.4 Filter media
10.5 Centrifugal filtration
10.5.1 Centrifugal filters
10.5.2 Filtration rates in centrifuges
10.6 Interceptive filtration
10.6.1 Deep bed filtration
10.6.2 Impingement filtration of gases
10.6.3 Interception mechanisms
10.6.4 Lamellar plate separators
11 Membrane filtration
11.1 Introduction
11.2 Membrane selection
11.3 Membrane filtration processes
11.4 Flux equations and selectivity
11.4.1 Flux definitions
11.4.2 Permeability for diffusion in porous membranes
11.4.3 Permeability for solution-diffusion in dense membranes
11.4.4. Selectivity and retention
11.5 Concentration polarization
11.6 Membrane modules
11.6.1 Design procedure
11.6.2 Solution
11.7 Concluding remarks
12 Separation method selection
12.1 Introduction
12.1.1 Industrial separation processes
12.1.2 Factors influencing the choice of a separation process
12.1.2.1 Economics
12.1.2.2 Feasibility
12.1.2.3 Product stability
12.1.2.4 Design reliability
12.2 Selection of feasible separation processes
12.2.1 Classes of processes
12.2.2 Initial screening
12.2.3 Separation factor
12.3 Separation of homogeneous liquid mixtures
12.3.1 Strengths of distillation
12.3.2 Limitations of distillation
12.3.3 Low relative volatilities
12.3.4 Overlapping boiling points
12.3.5 Low concentrations
12.3.6 Dissolved solids
12.4 Separaton systems for gas mixtures
12.4.1 General selection considerations
12.4.2 Comparison of gas separation
12.4.2.1 Absorption
12.4.2.2 Adsorption
12.4.2.3 Membrane separation
12.5 Separation methods for solid-liquid mixtures
Appendix
Answers to exercises
References and further reading
Index

Citation preview

De Gruyter Graduate De Haan · Bosch Industrial Separation Processes

Also of Interest Industrial Chemistry Benvenuto, 2014 ISBN 978-3-11-029589-4, e-ISBN 978-3-11-029590-0

Engineering Catalysis Murzin, 2013 ISBN 978-3-11-028336-5, e-ISBN 978-3-11-028337-2

Micro Process Engineering - Explained Fundamentals, Devices, Applications Dittmeyer, Brandner, Kraut, Pfeifer, 2013 ISBN 978-3-11-026538-5, e-ISBN 978-3-11-026635-1

Biorefinery From Biomass to Chemicals and Fuels Aresta, Dibenedetto, Dumeignil (Eds.), 2012 ISBN 978-3-11-026023-6, e-ISBN 978-3-11-026028-1

Green Processing and Synthesis Hessel, Volker (Editor-in-Chief) ISSN 2191-9542, e-ISSN 2191-9550

Reviews in Chemical Engineering Luss, Dan/Brauner, Neima (Editors-in-Chief) ISSN 0167-8299, e-ISSN 2191-0235

André B. de Haan · Hans Bosch

Industrial Separation Processes Fundamentals

DE GRUYTER

Authors Prof. Dr. Ir. André B. de Haan Eindhoven University of Technology Dpt. of Chemical Engineering and Chemistry PO Box 513 5600 MB Eindhoven Netherlands [email protected]

Dr. Ir. Hans Bosch University of Twente Faculty of Science & Technology Drienerlolaan 5 7522 NB Enschede Netherlands [email protected]

This book has 222 figures and 18 tables.

ISBN 978-3-11-030669-9 e-ISBN 978-3-11-030672-9 Library of Congress Cataloging-in-Publication Data A CIP catalog record for this book has been applied for at the Library of Congress. Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available in the Internet at http://dnb.dnb.de. © 2013 Walter de Gruyter GmbH, Berlin/Boston. The publisher, together with the authors and editors, has taken great pains to ensure that all information presented in this work (programs, applications, amounts, dosages, etc.) reflects the standard of knowledge at the time of publication. Despite careful manuscript preparation and proof correction, errors can nevertheless occur. Authors, editors and publisher disclaim all responsibility and for any errors or omissions or liability for the results obtained from use of the information, or parts thereof, contained in this work. Typesetting: META Systems GmbH, Wustermark Printing and binding: Hubert & Co. GmbH & Co. KG, Göttingen ♾ Printed on acid-free paper Printed in Germany www.degruyter.com

Preface

This textbook is written originally to support the graduate course “Separation Technology” at the University of Twente and has been later also used at Eindhoven University of Technology and other institutions of higher education. Our main objective is to present an overview of the fundamentals underlying the most frequently used industrial separation methods. We focus on their physical principles and the basic computation methods that are required to assess their technical and economical feasibility for a given application. Thus, design calculations are limited to those required for the development of conceptual process schemes. To keep computational time within the limits of our course structure, most examples and exercises of homogeneous mixtures are limited to binary mixtures. The textbook is organized into three main parts. Separation processes for homogeneous mixtures are treated in the parts on equilibrium based molecular separations (Chapters 2–5) and rate-controlled molecular separations (Chapters 6–8). The part on mechanical separation technology presents in Chapters 9–11 an overview of the most important techniques for the separation of heterogeneous mixtures. In each chapter a short overview is given of the most commonly used equipment types. Only for gas-liquid contactors Chapter 4 goes into more detail about their design an operation because they are the most commonly used industrial contactors. Unique is Chapter 12, which considers the selection of an appropriate separation process for a given separation task. The design of separation processes can only be learned by doing. The most important aspects are the combination of the right material balances with the thermodynamic equilibrium relations and mass transport equations. To support you in applying the presented material, we have included exercises at the end of each chapter. Short answers are given at the end of this book; detailed solutions are given in a separate solution manual that is available at http://dx.doi.org/ 10.1515/9783110306729_suppl. In the preparation of this book, a number of colleagues provided valuable comments and suggestions. Among these, we wish to acknowledge Wytze Meindersma (Eindhoven University of Technology), Boelo Schuur (University Twente) and Nieck Benes (University Twente). May, 2013

André B. de Haan Hans Bosch

Contents

1 Characteristics of separation processes

1.1

Significance of separations . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1

1.2 1.2.1 1.2.2 1.2.3

Characteristics of separation Categorization . . . . . . . . Separating agents . . . . . . Separation factors . . . . . .

. . . .

4 4 4 6

1.3 1.3.1 1.3.2

Industrial separation methods . . . . . . . . . . . . . . . . . . . . . . . . . . Exploitable properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Important molecular separations . . . . . . . . . . . . . . . . . . . . . . . . .

7 7 9

1.4 1.4.1 1.4.2

Inherent selectivities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Equilibrium based processes . . . . . . . . . . . . . . . . . . . . . . . . . . . Rate controlled processes . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

11 11 12

processes . . . . . . . . . . . . . . . . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

2 Evaporation and distillation

2.1 2.1.1 2.1.2

Separation by evaporation . . . . . . . . . . . . . . . . . . . . . . . . . . . . Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Vapor-liquid equilibria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

15 15 16

2.2 2.2.1 2.2.2 2.2.2.1 2.2.2.2

Separation by single-stage partial Differential distillation . . . . . . Flash distillation . . . . . . . . . . Specifying T and L / V . . . . . . . Specifying T and ptot . . . . . . .

evaporation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

21 21 22 24 25

2.3 2.3.1 2.3.2 2.3.3 2.3.4

Multistage distillation . . . . . . Distillation cascades . . . . . . . Column distillation . . . . . . . Feasible distillation conditions . External column balances . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

26 26 27 29 29

2.4 2.4.1 2.4.1.1 2.4.1.2 2.4.1.3 2.4.1.4 2.4.2 2.4.2.1 2.4.2.2

McCabe-Thiele analysis . . . . . . . . . . . . . . Internal balances . . . . . . . . . . . . . . . . . . Rectifying section . . . . . . . . . . . . . . . . . Stripping section . . . . . . . . . . . . . . . . . . Feed stage considerations . . . . . . . . . . . . . Feed line . . . . . . . . . . . . . . . . . . . . . . Required number of equilibrium stages . . . . Graphical determination of stages and location Limiting conditions . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . of feed stage . . . . . . . .

. . . . . . . . .

. . . . . . . . .

. . . . . . . . .

. . . . . . . . .

. . . . . . . . .

. . . . . . . . .

. . . . . . . . .

. . . . . . . . .

30 30 31 32 33 35 36 36 37

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

viii

Contents

2.4.2.3 2.4.2.4 2.4.3

Fenske and underwood equations . . . . . . . . . . . . . . . . . . . . . . . . Use of murphree efficiency . . . . . . . . . . . . . . . . . . . . . . . . . . . Energy requirements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

40 42 43

2.5 2.5.1 2.5.2 2.5.3

Advanced distillation techniques . . . . . . . . . . . Batch distillation . . . . . . . . . . . . . . . . . . . . . Continuous separation of multiple product mixtures Separation of azeotropes . . . . . . . . . . . . . . . .

. . . .

44 44 45 46

3.1

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

53

3.2

The aim of absorption . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

55

3.3

General design approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

56

3.4 3.4.1 3.4.2

Absorption and stripping equilibria . . . . . . . . . . . . . . . . . . . . . . . Gaseous solute solubilities . . . . . . . . . . . . . . . . . . . . . . . . . . . . Minimum absorbent flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

57 57 59

3.5 3.5.1 3.5.2 3.5.3

Absorber and stripper design Operating lines for absorption Stripping analysis . . . . . . . Analytical kremser solution .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

60 60 64 65

3.6 3.6.1 3.6.2 3.6.3 3.6.4

Industrial absorbers . . . . . . . . . Packed columns . . . . . . . . . . . Plate columns . . . . . . . . . . . . Spray and bubble columns . . . . . Comparison of absorption columns

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

68 69 70 70 71

4.1

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

77

4.2

Modeling mass-transfer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

77

4.3 4.3.1 4.3.2 4.3.3

Plate columns . . . . . . . . Dimensioning a tray column Height of a tray column . . Diameter of a tray column .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

83 85 86 91

4.4 4.4.1 4.4.2 4.4.3 4.4.4 4.4.5 4.4.6

Packed columns . . . . . . . . . Random packing . . . . . . . . . Structured packing . . . . . . . . Dimensioning a packed column Height of a packed column . . . Minimum column diameter . . Pressure drop . . . . . . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

97 98 99 100 101 104 106

4.5

Criteria for column selection . . . . . . . . . . . . . . . . . . . . . . . . . . .

107

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

3 Absorption and stripping

. . . .

. . . .

4 General design of gas/liquid contactors

. . . .

Contents

ix

5 Liquid-liquid extraction

5.1 5.1.1 5.1.2 5.1.3

Liquid-liquid extraction . Introduction . . . . . . . Liquid-liquid equilibria . Solvent selection . . . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

111 111 113 115

5.2 5.2.1 5.2.2 5.2.3 5.2.4

Extraction schemes . . . Single equilibrium stage Cocurrent cascade . . . . Crosscurrent . . . . . . . Countercurrent . . . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

117 117 118 119 120

5.3 5.3.1 5.3.2 5.3.3 5.3.4

Design of countercurrent extractions . . . . . . . . . . . . . Graphical McCabe-Thiele method for immiscible systems Analytical kremser method . . . . . . . . . . . . . . . . . . . Graphical method for partial miscible systems . . . . . . . Efficiency of an ideal non-equilibrium mixer . . . . . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

121 121 124 124 129

5.4 5.4.1 5.4.2 5.4.3 5.4.4 5.4.5

Industrial liquid-liquid extractors Mixer-settlers . . . . . . . . . . . . Mechanically agitated columns . Unagitated and pulsed columns . Centrifugal extractors . . . . . . . Selection of an extractor . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

130 130 131 132 134 135

6.1

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

143

6.2 6.2.1 6.2.2 6.2.3

Adsorption fundamentals Industrial adsorbents . . Equilibria . . . . . . . . . Kinetics . . . . . . . . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

145 145 150 154

6.3 6.3.1 6.3.2 6.3.3

Fixed-Bed adsorption . . . . . . . . . . Bed profiles and breakthrough curves Equilibrium theory model . . . . . . . Modeling of mass transfer effects . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

156 156 158 161

6.4 6.4.1 6.4.2 6.4.3

Basic adsorption cycles . . . . . . . . Temperature-swing . . . . . . . . . . Pressure-swing . . . . . . . . . . . . . Inert and displacement purge cycles

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

162 164 165 166

6.5 6.5.1 6.5.2

Principles of ion exchange . . . . . . . . . . . . . . . . . . . . . . . . . . . . Ion exchange resins . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Equilibria and selectivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

167 167 169

6.6

Ion exchange processes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

170

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

6 Adsorption and ion exchange

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

7 Drying of solids

7.1

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

177

7.2 7.2.1

Humidity definitions of carrier gas . . . . . . . . . . . . . . . . . . . . . . . Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

178 178

x

Contents

7.2.2 7.2.3

System air-water: a special case . . . . . . . . . . . . . . . . . . . . . . . . . Psychrometric charts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

180 182

7.3 7.3.1

Moisture in solids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Bound and unbound water . . . . . . . . . . . . . . . . . . . . . . . . . . . .

183 184

7.4 7.4.1 7.4.2 7.4.3

Drying mechanisms . . . . Constant drying rate . . . . Falling drying rate . . . . . Estimation of drying time

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

188 189 190 190

7.5 7.5.1 7.5.2 7.5.3

Classification of drying operations Direct-Heat dryers . . . . . . . . . . Contact dryers . . . . . . . . . . . . Other drying methods . . . . . . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

192 194 199 201

8.1

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

207

8.2 8.2.1 8.2.2

Crystal characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Morphology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Crystal size distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

209 210 211

8.3 8.3.1 8.3.2 8.3.3 8.3.4 8.3.5

Crystallization operating modes from solutions Operating modes . . . . . . . . . . . . . . . . . . Cooling crystallizers . . . . . . . . . . . . . . . Evaporating and vacuum crystallizers . . . . . Continuous crystallizers . . . . . . . . . . . . . Basic yield calculations . . . . . . . . . . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

212 212 214 215 217 219

8.4 8.4.1 8.4.2 8.4.3 8.4.4

Crystallizer modeling and design . . . . . . . . . . . . . . . Supersaturation and metastability . . . . . . . . . . . . . . . Nucleation . . . . . . . . . . . . . . . . . . . . . . . . . . . . Crystal growth . . . . . . . . . . . . . . . . . . . . . . . . . . Population balance equations for the MSMPR crystallizer

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

221 221 223 226 229

8.5 8.5.1 8.5.2

Other crystallization techniques . . . . . . . . . . . . . . . . . . . . . . . . . Precipitation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Melt crystallization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

232 232 233

. . . .

. . . .

. . . .

. . . .

8 Crystallization and precipitation

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

9 Sedimentation and settling

9.1

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

241

9.2 9.2.1 9.2.2 9.2.3 9.2.4 9.2.5

Gravity sedimentation . . . . . . . . . . . . . . . . . . . Sedimentation mechanisms . . . . . . . . . . . . . . . . Dilute sedimentation . . . . . . . . . . . . . . . . . . . Hindered settling . . . . . . . . . . . . . . . . . . . . . . Continuous sedimentation tank (gravity settling tank) Gravity sedimentation equipment . . . . . . . . . . . .

. . . . . .

242 242 245 247 248 250

9.3 9.3.1

Centrifugal sedimentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . Particle velocity in a centrifugal field . . . . . . . . . . . . . . . . . . . . .

251 251

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

Contents

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

xi

9.3.2 9.3.3 9.3.4 9.3.5 9.3.6

Sedimenting centrifuges . . . . . . . . Bowl centrifuge separation capability . The sigma concept . . . . . . . . . . . Capacity of disc centrifuges . . . . . . Hydrocyclones . . . . . . . . . . . . . .

. . . . .

. . . . .

252 254 256 258 259

9.4 9.4.1 9.4.2

Electrostatic precipitation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Principles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Equipment and collecting efficiency . . . . . . . . . . . . . . . . . . . . . .

263 263 265

10 Filtration

10.1

The filtration processfiltration process . . . . . . . . . . . . . . . . . . . . .

271

10.2 10.2.1 10.2.2 10.2.3 10.2.4

Filtration fundamentals . . . . . . . . . . . . . Flow through packed beds . . . . . . . . . . . Cake filtration . . . . . . . . . . . . . . . . . . Constant pressure and constant rate filtration Compressible cakes . . . . . . . . . . . . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

274 274 275 277 278

10.3 10.3.1 10.3.2 10.3.3

Filtration equipment . . . . . . . . . . . Continuous large-scale vacuum filters Batch vacuum filters . . . . . . . . . . Pressure filters . . . . . . . . . . . . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

279 279 283 284

10.4

Filter media . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

286

10.5 10.5.1 10.5.2

Centrifugal filtration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Centrifugal filters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Filtration rates in centrifuges . . . . . . . . . . . . . . . . . . . . . . . . . .

287 287 289

10.6 10.6.1 10.6.2 10.6.3 10.6.4

Interceptive filtration . . . . . . Deep bed filtration . . . . . . . Impingement filtration of gases Interception mechanisms . . . . Lamellar plate separators . . . .

. . . . .

292 292 293 295 297

11.1

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

301

11.2

Membrane selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

302

11.3

Membrane filtration processes . . . . . . . . . . . . . . . . . . . . . . . . . .

304

11.4 11.4.1 11.4.2 11.4.3 11.4.4.

Flux equations and selectivity . . . . . . . . . . . . . . . . Flux definitions . . . . . . . . . . . . . . . . . . . . . . . . Permeability for diffusion in porous membranes . . . . . Permeability for solution-diffusion in dense membranes . Selectivity and retention . . . . . . . . . . . . . . . . . . .

. . . . .

307 307 309 310 312

11.5

Concentration polarization . . . . . . . . . . . . . . . . . . . . . . . . . . . .

313

. . . . .

. . . . .

. . . . .

. . . . .

. . . .

. . . . .

. . . .

. . . . .

. . . .

. . . . .

. . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

11 Membrane filtration

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

xii

Contents

11.6 11.6.1 11.6.2

Membrane modules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Design procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

316 320 320

11.7

Concluding remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

321

12 Separation method selection

12.1 12.1.1 12.1.2 12.1.2.1 12.1.2.2 12.1.2.3 12.1.2.4

Introduction . . . . . . . . . . . . . Industrial separation processes . . . Factors influencing the choice of a Economics . . . . . . . . . . . . . . Feasibility . . . . . . . . . . . . . . . Product stability . . . . . . . . . . . Design reliability . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . separation process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

325 325 327 327 328 329 329

12.2 12.2.1 12.2.2 12.2.3

Selection of feasible separation processes Classes of processes . . . . . . . . . . . . . Initial screening . . . . . . . . . . . . . . . Separation factor . . . . . . . . . . . . . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

330 330 330 332

12.3 12.3.1 12.3.2 12.3.3 12.3.4 12.3.5 12.3.6

Separation of homogeneous liquid mixtures Strengths of distillation . . . . . . . . . . . . Limitations of distillation . . . . . . . . . . . Low relative volatilities . . . . . . . . . . . . Overlapping boiling points . . . . . . . . . . Low concentrations . . . . . . . . . . . . . . Dissolved solids . . . . . . . . . . . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

333 333 335 337 340 341 342

12.4 12.4.1 12.4.2 12.4.2.1 12.4.2.2 12.4.2.3

Separaton systems for gas mixtures . General selection considerations . . . Comparison of gas separation . . . . Absorption . . . . . . . . . . . . . . . Adsorption . . . . . . . . . . . . . . . Membrane separation . . . . . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

343 343 344 344 345 346

12.5

Separation methods for solid-liquid mixtures . . . . . . . . . . . . . . . . .

346

. . . . . .

. . . . . .

. . . . . .

. . . . . .

Appendix

Answers to exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References and further reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

351 359 363

1 Characteristics of separation processes

1.1 Significance of separations When ethanol is placed in water, it dissolves and tends to form a solution of uniform composition. There is no simple way to separate the ethanol and the water again. This tendency of substances to form a mixture is a spontaneous, natural process that is accompanied by an increase in entropy or randomness. In order to separate the obtained mixture into its original species, a device, system or process must be used that supplies the equivalent of thermodynamic work to induce the desired separation. The fact that naturally occurring processes are inherently mixing processes makes the reverse procedure of “unmixing” or separation processes one of the most challenging categories of engineering problems. We can now define separation processes as: Those operations that transform a mixture of substances into two or more products that differ from each other in composition

The separation of mixtures, including enrichment, concentration, purification, refining and isolation are of extreme importance to chemists and chemical engineers. Separation technology has been practiced for millennia in the food, material processing, chemical and petrochemical industry. In the food, material processing and petrochemical industry most processes do not involve a reaction step but are merely used for the recovery and purification of products from natural resources. Examples are oil refining (Fig. 1.1), metal recovery from ores and the isolation and purification of sugar from sugar beats or sugar cane schematically shown in Fig. 1.2. In this process a combination of separation techniques (washing, extraction, pressing, drying, clarification, evaporation, crystallization, filtration) are used to isolate the desired product, sugar, in its right form and purity. Within the chemical industry separation technology is mostly used in conjunction with chemical reactors. One of the main characteristics of all these chemical processes is that many more separation steps are used compared to the amount of chemical reaction steps. An illustrative example is the oxidation of ethylene to ethylene oxide shown in Fig. 1.3. After the reaction several absorption, desorption and distillation steps are required to obtain the ethylene oxide as a pure product. The above examples serve to illustrate the importance of separation operations in the majority of industrial chemical processes. Looking at the flow sheets in

2

1 Characteristics of separation processes

Fig. 1.1 Very simple schematic of a refinery for converting crude oil into products.

Fig. 1.2 Processing sequence for producing sugar from sugar beets.

Figs. 1.2 and 1.3 it should not be surprising that separation processes account for 40–90 % of capital and operating costs in industry and their proper application can significantly reduce costs and increase profits. As illustrated by Fig. 1.4, separation operations are employed to serve a variety of functions:

1.1 Significance of separations

Fig. 1.3 Ethylene oxide production by oxygen oxidation.

Fig. 1.4 Schematic of the functions of separations in manufacturing.

3

4

• • • • • •

1 Characteristics of separation processes

Removal of impurities from raw materials and feed mixtures Recycling of solvents and unconverted reactants Isolation of products for subsequent purification or processing Purification of products, product classes and recycle streams Recovery and purification of by-products Removal of contaminants from air and water effluents

1.2 Characteristics of separation processes 1.2.1 Categorization There are many examples where a separation of a heterogeneous feed consisting of more than one phase is desired. In such cases it is often beneficial to first use some mechanical means based on gravity, centrifugal force, pressure reduction, or an electric and/or magnetic field to separate the phases. Such processes are called mechanical separation processes. For example, a filter of a centrifuge serves to separate solid and liquid phases from slurry. Vapor-liquid separators segregate vapor from liquid. Then, appropriate separation techniques can be applied to each phase. Most other separation processes deal with single, homogeneous mixtures (solid, liquid or gas) and involve a diffusion transfer of material from the feed stream to one of the product streams. Often a mechanical separation is employed to separate the product phases in one of these processes. These operations are referred to as molecular separation processes. Most molecular separation processes operate through equilibration of two immiscible phases, which have different compositions at equilibrium. Examples are evaporation, absorption, distillation and extraction processes. We shall call these equilibrium-based processes. In another class of molecular separation processes the efficiency is mainly controlled by the transport rate through media resulting from a gradient in partial pressure, temperature, composition, electric potential or the like. They are called rate-controlled processes such as adsorption, ion exchange, crystallization and drying. Important in their design is the factor time that determines in combination with the transport rates their optimal, usually cyclic way of operation. 1.2.2 Separating agents A simplified scheme of a separation process is shown in Fig. 1.5. The feed mixture may consist of one or several streams and can be vapor, liquid or solid. From the fundamental nature of separation, there must be at least two product streams, which differ from each other in composition. If the feed mixture is a homogeneous solution a second phase must be developed before separation of chemical species can be achieved. This second phase can be created by an energy-

1.2 Characteristics of separation processes

5

Fig. 1.5 General separation process.

Fig. 1.6 General separation techniques (a) by phase creation; (b) solvent addition; (c) solid mass separating agent; (d) by barrier; (e) by force of field gradient.

separating agent (ESA), mass-separating agent (MSA), and barrier or external fields as shown in Fig. 1.6. The most common industrial technique, Fig. 1.6a, is the application of an energy-separating agent to create a second phase (vapor, liquid, or solid) that is immiscible with the feed phase. Such applications of an energy-separating agent involve energy (heat) transfer to or from the mixture. For example, in evapora-

6

1 Characteristics of separation processes

tion the separating agent is the heat (energy) supplied that causes the formation of a second (vapor) phase. A vapor phase may also be created by pressure reduction. A second technique is to create a second phase in the system by the introduction of a mass-separating agent. Mass separating agents may be used in the form of a liquid (absorption, extractive distillation, extraction) or a solid (adsorption, ion exchange). In extraction processes, Fig. 1.6b, the separating agent is a solvent that selectively dissolves some of the species in the feed mixture. Of growing importance are techniques that involve the addition of solid particles, Fig. 1.6c, which selectively adsorb solutes on their internal surface to create separation. Some separation processes utilize more than one separating agent. An example is extractive distillation, where a mixture of components with close boiling points is separated by adding a solvent (mass separating agent), which serves to enhance the volatility of some components stronger than others, and heat (energy separating agent) in a distillation scheme to generate a more volatile product and a less volatile product. Less common is the use of a barrier, Fig. 1.6d, which restricts the movement of certain chemical species with respect to other species. External fields, Fig. 1.6e, of various types are sometimes applied for specialized separations. For all of the general techniques of Fig. 1.6, the separations are achieved by enhancing the rate of mass transfer of certain species relative to all species. The driving force and direction of mass transfer by diffusion is governed by thermodynamics, with the usual limitations of equilibrium. Thus both transport and thermodynamic considerations are crucial in the design of separation operations. Mass transfer governs the rate of separation, while the extent of separation is limited by thermodynamic equilibrium. 1.2.3 Separation factors An important consideration in the selection of feasible separation methods is the degree of separation that can be obtained between two key components of the feed. Since the object of a separation device is to produce products of differing compositions, it is logical to define this degree of separation in terms of product compositions: This is commonly done through the separation factor for the separation of component A from component B between phases1 and 2, defined for a single stage of contacting as: CA,1 CA,2 SFA,B =

CB,1 CB,2

xA,1 xA,2 =

xB,1 xB,2

(1.1)

where C and x are composition variables, such as mole fraction, mass fraction or concentration. The value of the separation factor is limited by thermodynamic equilibrium, except in the case of membrane separations that are controlled by

1.3 Industrial separation methods

7

relative rates of mass transfer through the membrane. An effective separation is accomplished when the separation factor is significantly different from unity. If SF > 1 component A tends to concentrate in the product 1 while component B accumulates in product 2. On the other hand, if SF < 1, component B tends to concentrate preferentially in product 1. In general components A and B are designated in such a manner that the separation factor is larger than unity. Consequently, the larger the value of the separation factor is, the more feasible the particular operation. In real processes the separation factor can reflect the differences in equilibrium compositions and transport rates as well as the construction and flow configuration of the separation device. For this reason it is convenient to define in section 1.4 an inherent separation factor, the selectivity, which would be obtained under idealized conditions. For equilibrium separation processes the selectivity corresponds to those product compositions, which will be obtained when simple equilibrium is attained between the product phases. For rate-controlled separation processes the selectivity represents those product compositions, which will occur in the presence of the underlying physical transport mechanism alone. Complications from competing transport phenomena, flow configurations, or other extraneous effects are excluded.

1.3 Industrial separation methods 1.3.1 Exploitable properties The objective of the design of a separation process is to exploit property differences in the most economical manner to obtain the required separation factors. Mechanical separations exploit the property differences between the coexisting phases by applying a certain force to the heterogeneous mixture. Table 1.1 presents an overview of the different mechanical separations classified according to the principle involved. The achieved extend of separation in molecular separations depends on the exploitation of differences in molecular, thermodynamic and transport properties of the different chemical species present in the feed. Some important bulk thermodynamic and transport properties are: Vapor pressure, Adsorptivity, Solubility and Diffusivity. These differences in bulk properties result from differences in properties of the molecules themselves such as: Molecular weight Van der Waals volume Van der Waals area Molecular shape Dipole moment

Polarizability Dielectric constant Electric charge Radius of gyration

Table 1.2 indicates the importance of the main molecular properties in determining the value of the separation factor for various separation processes. Classifica-

8

1 Characteristics of separation processes

Table 1.1

Separation principles of mechanical separations.

Technique

Applied Mechanical Force

Technique

Applicable Separation Principle for particles (micron)

Sedimentation Ch 9

Gravity

Settlers Classifiers Centrifuges Cyclones Decanters Electrostatic precipitators Liquid + Solid

>100

Density difference

1–1000

Density difference

0.01–10

Charge on fine solid particles Attraction by Magnetism

Sieves Filtration Filtration Presses Sieves Membranes Centrifuges

>100

Centrifugal

Electrostatic Magnetic Filtration Ch 10, Ch 11

Gravity Pressure

Centrifugal Impingement

Table 1.2

Particle size larger than pore size of filter medium 0.001−1000 Particle size larger than pore size of filter medium

1–1000

Filters 0.1–1000 Scrubbers Impact separators

Particle size larger than pore size of filter medium Size difference

Relation of the separation factor to difference in molecular properties.

Distillation Crystallization Extraction Absorption Adsorption Membrane Filtration Ion Exchange Drying

Molecular weight

Molecular size

Molecular shape

Dipole Molecular moment and charge Polarizability

Chemical Reaction

++ + + + + −

+ ++ + + ++ ++

− ++ + ++ ++ +++

++ + ++ +++ ++ −

− ++ − − − −

− − ++ ++ −

− ++

− +−

− −

− +

− −

+++ −

tion of separation processes in terms of the molecular properties that primarily govern the separation factor can be quite useful for the selection of candidate processes for separating a given mixture. Processes that emphasize molecular properties in which the components differ to the greatest should be given special

1.3 Industrial separation methods

9

attention. Although this categorization only provides general guidelines, the basic differences in the importance of different molecular properties in determining the separation factors for different separation processes are apparent from Table 1.2. For example, the separation factor in distillation reflects vapor pressures, which in turn reflect primarily the strength of intermolecular forces. The separation factor in crystallization, on the other hand, reflects primarily the ability of molecules of different kinds to fit together, and simple geometric factors of size and shape become much more important. Values of these properties for many substances are available in handbooks, specialized reference books and journals. Many of them can also be estimated using computer-aided process simulation programs. When not available, these properties must be estimated or determined experimentally if a successful application of the appropriate separation operation(s) is to be achieved. 1.3.2 Important molecular separations An overview of the most commonly used industrial molecular separations divided into equilibrium and rate based separations is given in Table 1.3. The table shows the phases involved, the separating agent and the physical or chemical principle on which the separation is based. It should be noted that the feed to a separation unit usually consists of a single vapor, liquid or solid phase. If the feed comprises two or more coexisting phases, it should be considered to separate the feed stream first into two phases by some mechanical means and then sending the

Fig. 1.7 Technological and use maturities of separation processes.

10

1 Characteristics of separation processes

Table 1.3

Commonly used molecular separation methods.

Separation Method

Chapter

Phase of the Feed

Equilibrium Based Processes Flash 2 Liquid and/or vapor Distillation 2 Liquid and/or vapor Absorption 3 Vapor Stripping Extractive distillation Azetropic distillation Extraction

3

Liquid

5

Liquid and/or vapor Liquid and/or vapor Liquid

Rate Controlled Processes Gas adsorption 6 Vapor Liquid adsorption Ion exchange

Heat transfer Heat transfer Liquid absorbent Vapor stripping agent Liquid solvent and heat transfer Liquid entrainer and heat transfer Liquid solvent

Vapor + Liquid

Solid adsorbent

Gas + Solid

Liquid

Solid adsorbent

Liquid + Solid

6

Liquid

Ion exchanger

Liquid + Solid

Solid

Liquid solvent

Liquid + Solid

Solid

Heat transfer

Vapor + Solid

Liquid

Heat transfer

Liquid + Solid

7

Crystallization 8

Separation Principle

Difference volatility Vapor + Liquid Difference volatility Liquid + Vapor Difference volatility Vapor + Liquid Difference volatility Vapor + Liquid Difference volatility Vapor + Liquid Difference volatility Liquid + Liquid Difference solubility

6

Leaching Drying

Separation Agent Products

in in in in in in in

Difference in adsorbability Difference in adsorbability Difference in chemical reaction Difference in solubility Difference in volatility Difference in solubility or melting point

separated phases to different separation units. Table 1.3 is not intended to be complete but serves as an overview of the most widely applied industrial separations. Some separation operations are well understood and can be readily designed from a mathematical model and/or scaled up to a commercial size from laboratory data. This technological and use maturity has been analyzed and is graphically illustrated in Fig. 1.7 for the molecular separation technologies from Table 1.3. At the “technological asymptote” it is assumed that everything is known about the process and no further improvements are possible. If a process is used to the fullest extend possible, it has approached its “use asymptote”. As expected, the degree to which a separation operation is technologically mature correlates well

1.4 Inherent selectivities

11

with its commercial use. Those operations ranked near the top are frequently designed without the need for any laboratory of pilot-plant test. Operations near the middle usually require laboratory data, while operations near the bottom require extensive research before they can be designed. Distillation, the most frequently used process, is closer to its technological and use asymptotes than any other process.

1.4 Inherent selectivities 1.4.1 Equilibrium based processes For separation processes based upon the equilibration of immiscible phases the selectivity is always based on the equilibrium ratio of the components, defined as the ratio of the mole fraction (concentration) in phase 1 to the mole fraction in phase 2, at equilibrium: KA ≡

xA,1

(1.2)

xA,2

Introduction in the previous equation for the separation factor SF, defined in Eq. 1.1, provides the relation that defines the inherent selectivity SAB in terms of the mole fractions as well as equilibrium ratios: xA,1 xA,2

KA SAB =

KB

=

(1.3)

xB,1 xB,2

For processes based on equilibration between gas and liquid phases the selectivity can be related to vapor pressures P 0 and activity coefficients γ: pA = yA Ptot = γA xA PA0

or

KA ≡

yA xA

=

γA PA0 Ptot

(1.4)

For a mixture that forms nearly ideal liquid solutions Raoult’s law applies (γ = 1). Then, the inherent selectivity for vapor-liquid separation operations employing energy as the separating agent such as partial evaporation, partial condensation or distillation is given by: SAB = αAB =

PA0 PB0

(1.5)

12

1 Characteristics of separation processes

The inherent selectivity αAB in a vapor-liquid system is commonly called the relative volatility. In case of non-ideal solutions in vapor-liquid separation processes, the expressions for the K values of the key components must be corrected by the liquid-phase activity coefficients according to Eq. 1.4 and the relative volatility becomes: αAB =

γA PA0

(1.6)

γB PB0

Noting that thermodynamic equilibrium requires equal activities of the components in both phases, the distribution coefficient KA for equilibrium between two non-ideal immiscible liquid phases 1 and 2 follows from the introduction of the activity aA = γA xA in Eq. 1.2: aA,1 xA,2

γA,2

γA,1

xA,1 KA ≡

=

aA,2

=

γA,1

(1.7)

γA,2 Hence now the inherent selectivity, for liquid-liquid extraction commonly denoted as βAB , becomes: γA,2 γA,1 SAB = βAB =

(1.8)

γB,2 γB,1

1.4.2 Rate controlled processes In adsorption processes a dynamic equilibrium is established for the distribution of the solute between the fluid and the solid surface. In Chapter 6 is shown that at low amounts adsorbed, the isotherm should approach a linear form and the following form of Henry’s law: qA = HA pA

or

qA = HA′ cA

(1.9)

where q is the equilibrium loading or amount adsorbed per unit mass of adsorbent, p is the partial pressure of a gas and c is the concentration of the solute. The inherent selectivity of the adsorbent is determined similarly to the inherent selectivity defined in the previous section. For small amounts of adsorbed material the inherent adsorption selectivity simply corresponds to the ratio of the

Exercises

13

equilibrium Henry constants HA /HB (or HA′ /HB′ ) which equals the ratio of adsorption constants bA /bB (see Eq. 6.4): qA pA SAB =

qB pB

HA =

HB

=

qmax ⋅ bA qmax ⋅ bB

bA =

(1.10)

bB

The inherent selectivity for permeation through a barrier is defined in a similar way. By definition, the flux of a component through a membrane is proportional to its permeability, PM (see Chapter 11). Under ideal circumstances the inherent selectivity of permeation through a barrier is then estimated by the ratios of the permeabilities PM,A and PM,B : SAB =

PM, A

(1.11)

PM, B

Nomenclature a b cA CA,i KA p P0 Ptot PM Q SAB SF xA,i x, y αAB βAB γ

activity adsorption constant concentration of component A concentration of component A in phase i equilibrium ratio (Eq. 1.2) partial pressure vapor (saturation) pressure total pressure permeability equilibrium amount adsorbed inherent selectivity (Eq. 1.3) separation factor (Eq. 1.1) mole fraction of component A in liquid phase I mole fraction of A in liquid and vapor phase, respectively relative volatility (Eq. 1.5) (= SAB for ideal vapor – liquid systems) inherent selectivity in liquid-liquid extraction (Eq. 1.8) activity coefficient

[−] see Ch. 6 [mol m−3] [mol m−3] [−] [−] [N m−2] [N m−2] see Ch. 11 see Ch. 6 [−] [−] [−] [−] [−] [−] [−]

Exercises 1 Compare and discuss the advantages and disadvantages of making separations using an energy separating agent (ESA) versus using a mass separating agent (MSA) 2 The system benzene-toluene adheres closely to Raoult’s law. The vapor pressures of benzene and toluene at 121 °C are 300 and 133 kPa. Calculate the relative volatility.

14

1 Characteristics of separation processes

3 As a part of the life support system for spacecraft it is necessary to provide a means of continuously removing carbon dioxide from air. It is not possible to rely upon gravity in any way to devise a CO2-air separation process. Suggest at least two separation schemes, which could be suitable for continuous CO2 removal from air under zero-gravity conditions. 4 Gold is present in seawater to a concentration level between 10−12 and 10−8 weight fraction, depending upon the location. Briefly evaluate the potential for recovering gold economically from seawater. 5 Assuming that the membrane characteristics are not changed, will the product-water purity in a reverse-osmosis seawater desalination process increase, decrease or remain constant as the upstream pressure increases. 6 Propylene and propane are among the light hydrocarbons produced by thermal and catalytic cracking of heavy petroleum fractions. Although propylene and propane have close boiling points, they are traditionally separated by distillation. Because distillation requires a large number of stages and considerable reflux and boil up flow rates compared to the feed flow, considerable attention has been given to the possible replacement of distillation with a more economical and less energy-intensive option. Based on the given properties of both species, propose some alternative properties that can be exploited to enhance the selectivity of propylene and propane separation. What kind of separation processes are based on these alternative properties. Property

Propylene

Propane

Molecular weight Vd Waals volume (m3/kmol) Vd Waals area (m2/kmol) Acentric Factor Dipole moment (debye) Radius of gyration (m * 1010) Melting point (K) Boiling point (K) Critical temperature (K) Critical pressure (MPa)

42.08 0.0341 5.06 0.142 0.4 2.25 87.9 225.4 364.8 4.61

44.10 0.0376 5.59 0.152 0.0 2.43 85.5 231.1 369.8 4.25

2 Evaporation and distillation

2.1 Separation by evaporation 2.1.1 Introduction A large part of the separations of individual substances in a homogeneous liquid mixture or complete fractionation of such mixtures into their individual pure components is achieved through evaporative separations. Evaporative separations are based on the difference in composition between a liquid mixture and the vapor formed from it. This composition difference arises from differing effective vapor pressures, or volatilities, of the components in the liquid mixture. The required vapor phase is created by partial evaporation of the liquid feed through adding heat, followed by total condensation of the vapor. Due to the difference in volatility of the components the feed mixture is separated into two or more products whose compositions differ from that of the feed. The resulting condensate is enriched in the more volatile components, in accordance with the vaporliquid equilibrium for the system at hand. When a difference in volatility does not exist separation by simple evaporation is not possible. The basis for planning evaporative separations is knowledge of the vapor-liquid equilibrium. Technically, evaporative separations are the most mature separation operations. Design and operating procedures are well established. Only when vapor-liquid equilibrium or other data are uncertain is a laboratory and/or pilotplant study necessary prior to the design of a commercial unit. The most elemen-

Fig. 2.1 Laboratory distillation setup.

16

2 Evaporation and distillation

tary form is simple distillation in which the liquid mixture is brought to boiling, partially evaporated and the vapor formed separated and condensed to form a product. This technique is commonly used in the laboratory for the recovery and/ or purification of products after synthesis in an experimental setup as illustrated by Fig. 2.1. 2.1.2 Vapor-liquid equilibria It is clear that equilibrium distributions of mixture components in the vapor and liquid phase must be different if separation is to be made by evaporation. At thermodynamic equilibrium the compositions are called vapor-liquid equilibria (VLE) that may be depicted as in Fig. 2.2. For binary mixtures the effect of distribution of mixture components between the vapor and liquid phases on the thermodynamic properties is illustrated in Fig. 2.3. Fig. 2.3a shows a representative boiling point diagram with equilibrium compositions as functions of temperature at a constant pressure. Commonly the more volatile (low boiling) component is used to plot the liquid and vapor compositions of the mixture. The lower line is the liquid bubble point line, the locus of points at which a liquid on heating forms the first bubble of vapor. The upper line is the vapor dew point line, representing points at which a vapor on cooling forms the first drop of condensed liquid. The region between the bubble and dew point lines is the twophase region where vapor and liquid coexist in equilibrium. At the equilibrium temperature Teq the liquid with composition xeq is in equilibrium with vapor composition yeq. Fig. 2.3b displays a typical y–x diagram that is obtained by plotting the vapor composition that is in equilibrium with the liquid composition at a fixed pressure or a fixed temperature. At equilibrium, the concentration of any component i present in the liquid mixture is related to its concentration in the

Fig. 2.2 Schematic vapor-liquid equilibrium.

2.1 Separation by evaporation

17

Fig. 2.3 Isobaric vapor-liquid equilibrium diagrams: (a) dew and bubble point; (b) y-x diagram.

vapor phase by the equilibrium ratio as defined in Eq. 1.2, also called the distribution coefficient Ki : Ki ≡

yi

(2.1)

xi

where yi is the mole fraction of component i in the vapor phase and xi the mole fraction of component i in the liquid phase. The distribution coefficient is a function of temperature and pressure only, not of compositions. The more volatile components in a mixture will have the higher values of Ki , whereas less volatile components will have lower values of Ki . The key separation factor in distillation is the selectivity of a component relative to a reference component. This selectivity, in VLE usually referred to as relative volatility, is defined in Eq. 1.3 as: yi xi

Ki αij =

Kj

=

yi xi

(2.2)

where component i is the more volatile one. The higher the value of the relative volatility, the more easily components may be separated by distillation. In ideal systems the behavior of vapor and liquid mixtures obeys Daltonʼs and Raoultʼs laws. Daltonʼs law relates the concentration of a component present in an ideal gas or vapor mixture to its partial pressure: pi = yi Ptot

(2.3)

18

2 Evaporation and distillation

where pi is the partial pressure of component i in the vapor mixture and Ptot is the total pressure of the system given by the sum of the partial pressures of all components in the system: N

Ptot = ∑ pi

(2.4)

i=1

Raoultʼs law relates the partial pressure of a component in the vapor phase to its concentration in the liquid phase, xi : pi = xi Pi0

(2.5)

where Pi0 is the vapor pressure of pure component i at the system temperature (saturation pressure). Combining equations (2.3) and (2.5) yields: yi Ptot = xi Pi0

(2.6)

This results in the following relations between the pure component vapor pressure, Pi0, its distribution coefficient, Ki , and the relative volatility of two components in an ideal mixture, αij : Ki =

yi

=

xi

Pi0

αij =

and

Ki Kj

Ptot

=

Pi0 Pj0

(1.5)

The latter equation shows that for ideal systems the relative volatility is independent of pressure and composition. For a binary system with components A and B, where yA = 1 – yB and xA = 1 – xB , the relative volatility equation, Eq. 2.2, can be rearranged to give y 1−y

= α⋅

x 1−x

or

y=

αx 1 + (α − 1) x

(2.7)

where x and y represent the mole fractions of the more volatile component A. This equation is used to express the concentration of a component in the vapor as a function of its concentration in the liquid and relative volatility. This function is plotted in Fig. 2.4 for various values of relative volatility. When relative volatility increases, the concentration of the most volatile component in the vapor increases. When the relative volatility is equal to 1, the concentrations of the most volatile component in the liquid and vapor phases are equal and a vapor/liquid separation is not feasible. At very low concentrations, Eq. 2.7 reduces to a linear relation1, represented as tangents in Fig. 2.4: y = α⋅x 1

See also page 24.

(2.8)

2.1 Separation by evaporation

19

Fig. 2.4 Vapor-liquid equilibrium compositions as a function of relative volatility.

Since vapor pressures of components depend on temperature, equilibrium ratios are a function of temperature. However, the relative volatility considerably less sensitive to temperature changes because it is proportional to the ratio of the vapor pressures. In general the vapor pressure of the more volatile component tends to increase at a slower rate with increasing temperature than the less volatile component. Therefore, the relative volatility generally decreases with increasing temperature and increases with decreasing temperature. Most liquid mixtures are non-ideal and require Raoultʼs law to be modified by including a correction factor called the liquid phase activity coefficient: pi = γi xi Pi0

(1.4)

Although at high pressures, the vapor phase may also depart from ideal vapor mixture, the common approach in distillation is to assume ideal vapor behavior and to correct non-ideal liquid behavior with the liquid phase activity coefficient. The standard state for reference for the liquid phase activity coefficient is commonly chosen as γi → 1 for the pure component. The liquid phase activity coefficient is strongly dependent upon the composition of the mixture. Positive deviations from ideality (γi > 1) are more common when the molecules of different compounds are dissimilar and exhibit repulsive forces. Negative deviations (γi < 1) occur when there are attractive forces between different compounds that do not occur for either component alone. For non-ideal systems the relations for the distribution coefficient and relative volatility as given in Eq. 1.5 now become: Ki =

γi Pi0 Ptot

αij =

Ki Kj

=

γiPi0 γjPj0

(1.6)

20

2 Evaporation and distillation

In non-ideal systems the distribution coefficients and relative volatility are dependent on composition because of the composition dependence of the activity coefficients. When the activity coefficient of a specific component becomes high enough an azeotrope may be encountered, meaning that the vapor and liquid compositions are equal and the components cannot be separated by conventional distillation. Fig. 2.5 shows binary vapor-liquid composition (x-y), temperaturecomposition (T-x) and pressure-composition (P-x) diagrams for non-azeotrope,

Fig. 2.5 Types of binary (a) temperature-composition; (b) pressure-composition and (c) x-y phase diagrams for vapor-liquid equilibrium.

2.2 Separation by single-stage partial evaporation

21

minimum azeotrope and maximum azeotrope systems. A minimum boiling azeotrope boils at a lower temperature than each of the components in their pure states. When separating the components of this type of system by distillation, such as ethanol-water, the overhead product is the azeotrope. A maximum-boiling azeotrope boils higher than either component in their pure states and is the bottom product of distillation. An example of this type of system is acetone-chloroform. Vapor pressures for many compounds are published in literature (see references [29−34]) and often correlated as a function of temperature by the Antoine equation: ln Pi0 = Ai −

Bi T + Ci

(2.9)

where Ai , Bi and Ci are Antoine constants of component i and T represents the temperature. Unfortunately, in some reference books these data are given in nonSI units such as pressure in mmHg and, sometimes, temperature in °F. Then appropriate conversions are in order.

2.2 Separation by single-stage partial evaporation 2.2.1 Differential distillation Two main modes are utilized for single-stage separation by partial evaporation. The most elementary form is differential distillation, in which a liquid is charged

Fig. 2.6 Simple differential distillation.

22

2 Evaporation and distillation

to the still pot and heated to boiling. As illustrated by Fig. 2.6, the method has a strong resemblance with laboratory distillation, shown previously in Fig. 2.1. The vapor formed is continuously removed and condensed to produce a distillate. Usually the vapor leaving the still pot with composition yD is assumed to be in equilibrium with perfectly mixed liquid in the still at any instant. The distillate is richer in the more volatile components and the residual unvaporized bottoms are richer in the less volatile components. As the distillation proceeds, the relative amount of volatile components composition of the initial charge and distillate decrease with time. Because the produced vapor is totally condensed, yD = xD , there is only one single equilibrium stage, the still pot. Simple differential distillation is not widely used in industry, except for the processing of high valued chemicals in small production quantities or for distillations requiring regular sanitation. 2.2.2 Flash distillation Flash distillation is the continuous form of simple single-stage equilibrium distillation. In a flash a continuous feed is partially vaporized to give a vapor richer in the more volatile components than the remaining liquid. Because of the single stage, usually only a limited degree of separation is achieved. However, in some cases, such as the seawater desalination, the large volatility difference between the components results in complete separation. A schematic diagram of the equipment for flash distillation is shown in Fig. 2.7. The pressurized liquid feed is heated and either flashed adiabatically across a nozzle to a lower pressure or, without the valve, partially vaporized isothermally. Either way the created vapor phase is separated from the remaining liquid in a flash drum. A demister or

Fig. 2.7 Flash drum: adiabatic flash distillation with valve or isothermal flash without valve.

2.2 Separation by single-stage partial evaporation

23

entrainment eliminator is often employed to prevent liquid droplets from being entrained in the vapor. When properly designed, the system in the flash chamber behaves very close to an equilibrium stage (TV = TL) due to the intimate contact between liquid and vapor. The designer of a flash system needs to know the liquid and vapor compositions that correspond to the operating pressure and temperature in the flash drum. The design calculations2 are based on the mass balances for the balance envelope shown as a dashed line in Fig. 2.7 and on vapor-liquid equilibrium data; x, y and z represent the compositions (mole fractions) of the liquid, vapor and feed, respectively. In a binary system the two independent mass balances are the overall mass balance and the component balance for the more volatile component: F=V+L

and

Fz = Vy + Lx

(2.10)

By introducing the liquid vapor ratio L/V the latter two equations reduce to the dimensionless operating line equation: y=−

L V

x+ 1+

(

L V)

z

(2.11)

Equilibrium data may be represented by relations such as given in Eq. 2.1, which take the form in case of a binary system with components A and B: y = KA x

and

1 − y = KB (1 − x)

(2.12)

By combining these two the equilibrium line may be expressed equally well in terms relative volatility: y=

αx 1 + (α − 1) x

with

α=

KA KB

(2.7)

In case of ideal gas and liquid phases, the relative volatility α is the VL-equilibrium constant, also referred to as K (see, e.g., Eqs. 3.21 and 4.5) In case of an isothermal flash no valve is used while the specified temperature is set through the heat exchanger. Gibbsʼ phase rule3 states that in a binary twophase system two degrees of freedom exist, thus one more parameter can be chosen freely. Given the feed composition, z, one option to define the system would be to specify temperature and liquid-vapor ratio, another option being the specification of temperature and total pressure. In the following two sections we will show the choice of the relevant set of equations and how to solve these.

2 3

Estimation of the drum size, related to the liquid and vapor fluxes, is outside the scope of this text book. Degrees of freedom = number of components – number of phases + 2.

24

2 Evaporation and distillation

2.2.2.1 Specifying T and L/V In this case the total pressure is not known beforehand and, noting that KA and KB are inversely proportional to Ptot , Eq. 2.7 should be used as equilibrium line. Vapor and liquid compositions follow from solving Eqs. 2.7 and 2.11 with known values of α and L/V. The partial pressures are calculated from Eq. 2.5 and the total pressure from Eq. 2.4. A graphical solution is very convenient since the simultaneous solution of Eqs. 2.7 and 2.11 is represented by the intersection of the equilibrium curve and the operating line. Instead of the vapor-liquid ratio, often the fraction feed remaining liquid q and fraction feed vaporized 1 – q are utilized: q≡

L F

and

(1 − q) =

V F

(2.13)

The operating line equation now becomes y=−

q 1−q

x+

1 1−q

z

(2.14)

On the x-y diagram in Fig. 2.8, showing an example for a benzene-water separation, three straight operating lines and the equilibrium line4 according to Eq. 2.7

Fig. 2.8 Graphical determination of the equilibrium benzene mole fraction in the vapor and liquid from a flash drum for several (1 – q) values. 4

Almost a straight line because of the very low values of the benzene mole fraction in the mixture.

2.2 Separation by single-stage partial evaporation

25

are plotted. Each operating line represents a fraction of feed evaporated. The intersection points give the vapor and liquid compositions, yeq and x eq , leaving the flash drum as the fraction of feed evaporated (1 – q) varies from 0.001 to 0.002 to 0.005. 2.2.2.2 Specifying T and Ptot Setting temperature and total pressure defines the liquid-vapor system in the flash drum. This can also be seen in Fig. 2.3a where at a given T and Ptot the compositions of the coexisting phases, xeq and yeq , are fixed. In this case q (hence the slope of the operating line) is unknown beforehand, hence the solution procedure now differs from that outlined in the previous section. However, three boundary conditions can be defined: – the feed composition, z, must fall between x eq and yeq to assure that a vaporliquid equilibrium exists at the given temperature and pressure, – in a vapor-liquid system q is be bounded between zero and unity, and – the condition that the sum of mole fractions in each phase equals unity, but in the form ∑ xi = ∑ yi. Now, the usual numerical procedure is to firstly find q by elimination of x and y from the operating line (Eq. 2.11) and equilibrium line (Eq. 2.12), noting that at given T and Ptot the distribution coefficients Ki are known constants. In a multicomponent mixture yi = Ki · xi and the operating line equation becomes: xi Ki = −

q 1−q

xi +

zi 1−q

or

xi =

zi Ki + (1 − Ki) q

(2.15)

But yi = Ki xi =

Ki zi Ki + (1 − Ki) q

(2.16)

and combination of Eqs. 2.15 and 2.16 with the boundary condition ∑ xi = ∑ yi leads to zi (1 − Ki) ∑ K + (1 − K ) q = 0 i i i

(2.17)

This procedure for a multicomponent mixture thus leads to a single equation which can be solved for the one unknown q. Once q is found between zero and unity, all xi and yi are calculated from Eqs. 2.15 and 2.16.

26

2 Evaporation and distillation

Solving Eq. 2.17 for q in a binary system (zA = z and zB = 1 − z) and substitution of the result in Eq. 2.15 gives q= −z

KB 1 − KB

− (1 − z)

KA 1 − KA

and

x=

1 − KB KA − KB

(2.18)

Note that KA and KB are functions of temperature and total pressure. This expression for x can also be obtained by elimination of y and α from Eqs. 1.5, 2.6 and 2.7.

2.3 Multistage distillation 2.3.1 Distillation cascades Flash distillation is a very simple unit operation that in most cases produces only a limited amount of separation. Increased separation is possible in a cascade of flash separators that produces one pure vapor and one pure liquid product. Within the cascade the intermediate product streams are used as additional feeds. The liquid streams are returned to the previous flash drum, while the produced vapor streams are forwarded to the next flash chamber. Fig. 2.9 shows the resulting countercurrent cascade, so called because vapor and liquid streams go in opposite directions. The advantages of this cascade are that there are no intermediate products and the two end products can both be pure and obtained in high yield.

Fig. 2.9 Flash drum cascades.

2.3 Multistage distillation

Fig. 2.10

27

General flowchart of a plate distillation column.

Although a significant advance, this multiple flash drum system is seldom used industrially. Operation and design is easier if part of the top vapor stream is condensed and returned to the first stage, reflux, and if part of the bottom liquid stream is evaporated and returned to the bottom stage, boilup. This allows control of the internal liquid and vapor flow rates at any desired level by applying all of the heat required for the distillation to the bottom reboiler and do all the required cooling in the top condenser. Partial condensation of intermediate vapor streams and partial vaporization of liquid streams is achieved by heat exchange between all pairs of passing streams. This is most effectively achieved by building the entire system in a column instead of the series of individual stages shown in Fig. 2.9. Intermediate heat exchange is the most efficient with the liquid and vapor in direct contact. The final result is a much simpler and cheaper device, the distillation column shown in Fig. 2.10. 2.3.2 Column distillation Most commercial distillations involve some form of multiple staging in order to obtain better separation than is possible by single vaporization and condensation. It is the most widely used industrial method of separating liquid mixtures in the chemical process industry. As shown by Fig. 2.10, most multistage distillations are

28

2 Evaporation and distillation

continuously operated column-type processes separating components of a liquid mixture according to their different boiling points in a more volatile distillate and a less volatile bottoms or residue. The feed enters the column at the equilibrium feed stage. Vapor and liquid phases flow countercurrently within the mass transfer zone of the column where trays or packings are used to maximize interfacial contact between the phases. The section of column above the feed is called the rectification section and the section below the feed is referred to as the stripping section. Although Fig. 2.10 shows a distillation column equipped with only 10 sieve trays, industrial columns may contain more than 100. The liquid from a tray flows through a downcomer to the tray below and vapor flows upward through the holes in the sieve tray to the tray above. Intimate contact between the vapor and liquid phases is created as the vapor passes through the holes in the sieve tray and bubbles through the pool of liquid residing on the tray. The vapors moving up the column from equilibrium stage to equilibrium stage are increasingly enriched in the more volatile components. Similarly, the concentration of the least volatile components increases in the liquid from each tray going downward. The overhead vapor from the column is condensed to obtain a distillate product. The liquid distillate from the condenser is divided into two streams. Part of is withdrawn as overhead product (D) while the remaining distillate is refluxed (Lo) to the top tray to enrich the vapors. The reflux ratio is defined as the ratio of the reflux rate to the rate of product removal. The required heat for evaporation is added at the base of the column in a reboiler, where the bottom tray liquid is heated and partially vaporized to provide the vapor for the stripping section. Plantsize distillation columns usually employ external steam powered kettle or vertical thermosyphon type heat exchangers (Fig. 2.11). Both can provide the amount of heat transfer surface required for large installations. Thermosyphon reboilers are favored when the bottom product contains thermally sensitive compounds, only a

Fig. 2.11

Industrial reboilers: (a) kettle-type; (b) vertical thermosyphon-type.

2.3 Multistage distillation

29

small temperature difference is available for heat transfer or heavy fouling occurs. The vapor from the reboiler is sent back to the bottom tray and the remaining liquid is removed as the bottom product. 2.3.3 Feasible distillation conditions Industrial distillation processes are restricted by operability of the units, economic conditions and environmental constraints. An upper limit exists for feasible operating temperatures. One reason is the thermal stability of the species in the mixtures to be separated. Many substances decompose at higher temperatures and some species are not even stable at their normal boiling points. A second reason for a maximum temperature limitation is the means of heat supply. In most cases the required energy is supplied by condensing steam. The pressure of the available steam places an upper limit on temperature levels that can be achieved. Only in special cases (crude oil, sulfuric acid), higher temperatures may be realized by heating with hot oil or natural gas burners. With high-pressure steam the maximum attainable temperatures are limited to 300 °C, which can be raised to temperatures as high as 400 °C when other media are used. The temperature of the coolant for the overhead condenser dictates the lower limit of feasible temperatures in distillation columns. In most cases water is used, resulting to a minimum temperature in the column of 40 to 50 °C. In most cases temperature constraints can be met by selecting a proper operating pressure. Operating temperatures can be decreased by the use of a vacuum. It is technically feasible to operate distillation columns at pressures up to 2 mbar. Lower pressures are seldom used because of the high operating and capital costs of the vacuum-producing equipment. A column can be operated at higher pressure to increase the boiling point of low boiling mixtures. The upper limit for the operating pressure lies in the range of the critical pressures of the constituents. In addition to temperature and pressure the feasible number of theoretical stages of industrial columns is also limited. Only in exceptional cases commercial distillation columns are constructed that contain more than 100 stages. 2.3.4 External column balances Once the separation problem has been specified, the task of the engineer is to calculate the design variables. The first step is the development of mass balances around the entire column using the balance envelope shown by the dashed outline in Fig. 2.12. The aim of these balances is to calculate flow rates and compositions for the distillate and bottom products, D and B, for the given problem specification. Analogously to the flash calculations in section 2.2.2, the overall mass balance becomes: F=D+B

(2.19)

30

2 Evaporation and distillation

Fig. 2.12 Distillation column overall mass balance envelope.

In binary distillation the specification of xD and xB for the more volatile component fixes the distillate and bottoms flow rates D and B through the additional overall component material balance: xF F = xD D + xB B

(2.20)

2.4 McCabe-Thiele analysis 2.4.1 Internal balances If the molar heat of vaporization for any component has approximately the same value, condensation of 1 mole of vapor will vaporize 1 mole of liquid. Thus liquid and vapor flow rates tend to remain approximately constant as long as heat losses from the column are negligible and the pressure is uniform throughout the column. This simplified situation is closely approximated for many distillations; constant molar vapor and liquid flow may be assumed in each section of the distillation column, given a value for the distillate flow D. For constant molar flow

2.4 McCabe-Thiele analysis

31

conditions, it is not necessary to consider energy balances in either the rectifying or stripping sections. Only material balances and a vapor-liquid equilibrium curve are required. Based on these assumptions the McCabe-Thiele method provides a graphical solution to the material balances and equilibrium relationships inside the column. In this chapter the graphical McCabe-Thiele method is used to visualize the essential aspects of multistage distillation calculations required to understand the fundamentals behind modern, computer-aided design methods. In addition to the number of equilibrium stages N, also the required minimum number of stages Nmin, minimal reflux ratio Rmin and optimal stage for feed entry are derived. Finally the overall energy balances are derived to determine condenser and reboiler duties. 2.4.1.1 Rectifying section As illustrated by Fig. 2.13, the rectifying section extends from the top stage 1, to just above the feed stage f. For the indicated balance envelope the material balance for the more volatile component over the total condenser and a top portion of rectifying stages 1 to n is written as: Vn′ + 1 yn + 1 = L ′n xn + DxD

(2.21)

Since V ′ and L ′ are constant throughout the rectifying section, Eq. 2.21 can be rewritten as: yn + 1 =

L′

(V ′ )

xn +

D

(V ′ )

xD

(2.22)

Eq. 2.22 is the operating line in the rectifying section. It relates the concentrations yn+1 and xn of the two passing streams V ′ and L ′ in the column and thus represents the mass balances in the rectifying section. The slope of the operating line is L ′/V ′, which is constant and < 1 because in the rectifying section V ′ > L ′. The liquid entering the top stage is the external reflux rate, L ′, and its ratio to the distillate rate, L ′/D, is defined as the reflux ratio R. Since V ′ = L ′ + D, the slope of the operating line is readily related to the reflux ratio: L′ V′

=

L′ L′ + D

R

=

and

R +1

D V′

=

D D + L′

=

1

(2.23)

R +1

Substitution into Eq. 2.22 produces the most useful form of the operating line for the rectifying section: yn + 1 =

R

(R + 1)

xn +

1

(R + 1)

xD

(2.24)

32

2 Evaporation and distillation

Fig. 2.13 (a) Mass balance envelope; (b) operating line for the rectifying section.

As shown in Fig. 2.13, the operating line plots as a straight line in the y-x-diagram with an intersection at y = xD on the y = x line, for specified values of R and xD .

2.4.1.2 Stripping section The stripping section extends from the feed to the bottom stage. Analogous to the rectifying section a bottom portion of the stripping stages, including the partial reboiler and extending up from stage N to stage m + 1, located somewhere below the feed is considered. A material balance for the more volatile component over the envelope indicated in Fig. 2.14 results in: L ″m x m = V m″ + 1 ym + 1 + BxB

(2.25)

which by the constant-molar-flow conditions rearranges into the operating line representing the mass balance of the passing streams V ″ and L ″ in the stripping section: ym + 1 =

L″

(V ) ″

xm −

B

(V ″ )

xB

(2.26)

The slope of this operating line for the stripping section is seen to be >1 because L ″ > V ″. This is the reverse of the conditions in the rectifying section. For known values of L ″, V ″ and xB , Eq. 2.26 can be plotted as a straight line with an intersection at y = xB on the y = x line and a slope of L ″/V ″, as shown in Fig. 2.14.

2.4 McCabe-Thiele analysis

Fig. 2.14

33

(a) Mass balance envelope; (b) operating line for the stripping section.

2.4.1.3 Feed stage considerations Thus far we have not considered the feed to the column. In the rectifying and stripping section of the column the operating line represents the mass balances. When material is added or withdrawn from the column, the mass balances will change and the operating lines will have different slopes and intercepts. The addition of the feed increases the liquid flow in the stripping section, increases the vapor flow in the rectifying section, or does both. In the rectifying section the vapor flow is greater than the liquid flow rate. In the stripping section the liquid flow is greater than the vapor flow rate. Obviously the phase and temperature of the feed affect the vapor and liquid flow rates in the column. For instance, if the feed is liquid, the liquid flow rate below the feed stage must be greater than the liquid flow above the feed stage, L ″ > L ′. Similarly V ′ should be larger than V ″ if the feed is a vapor. These effects can be quantified by writing mass and energy balances around the feed stage shown schematically in Fig. 2.15a. The overall mass balance for the balance envelope is: F + V ″ + L′ = L″ + V ′

(2.27)

and the related balance over just the liquid phase reads: LF = L ″ − L ′ ≡ q ⋅ F

(2.28)

where q is defined as the liquid fraction in the feed at feed stage temperature, similar to the q-factor defined in the Flash Distillation section. Under the condition that the feed is at feed stage temperature, q represents the moles of the liquid flow in the stripping section that result from the introduction of each mole of feed. Assuming that the molar vapor enthalpies hV nor the molar liquid enthalpies hL vary much with composition, the energy balance across the feed stage equals:

34

2 Evaporation and distillation

Fig. 2.15 (a) Mass balance envelope and (b) q-line for the feed stage.

FhF + V ″hV + L ′ hL = L ″ hL + V ′hV

(2.29)

Rewriting the mass balance Eq. 2.27 in terms of V ″ − V ′ and substitution into the energy balance gives: FhF + (L ″ − L ′ ) hV − FhV = (L ″ − L ′ ) hL

(2.30)

which is easily rearranged to yield: q≡

L″ − L′ F

=

hV − hF hV − hL

(2.31)

Hence the liquid feed fraction equals the ratio of the heat required to completely evaporate a mole of feed hV – hF and the molar heat of evaporation hV – hL. Since the liquid and vapor enthalpies can be estimated, we can easily calculate q from Eq. 2.31 for the condition that the feed consists of a partially evaporated liquid and vapor mixture (1 > q > 0). The feed may also be a saturated liquid at its bubble point (q = 1), or a saturated vapor at its dew point (q = 0). The energy balance in Eq. 2.29 does not comprise a term to bring the feed to the feed stage temperature. Thus if the feed is introduced as cold liquid (q > 1) or as a superheated vapor (q < 0), the value of q calculated from Eq. 2.31 should be corrected for the heat involved5. The required correction is outside the scope of this text book but can be found in the literature references.

5 We will denote q-values subject to this temperature correction as q ′. For small temperature differences q ′ ≈ q.

2.4 McCabe-Thiele analysis

35

2.4.1.4 Feed line After quantifying the value for q, the changes in liquid and vapor flow rates due to the introduction of the feed on the feed stage are calculated. The contribution of the feed stream to the internal flow of liquid is q · F (Eq. 2.28), so the total flow rate of liquid and vapor in the stripping section become: L ″ = L ′ + qF

V ″ = V ′ − (1 − q) F

and

(2.32)

Although with these values for L ″ and V ″ Eq. 2.26 can be used to locate the stripping operating line it is more common to use an alternative method that involves the feed-line or q-line. The intersection point of the rectifying and stripping operating lines is derived by writing the mass balance equations for both sections for constant molar flow: V ′ y = L ′ x + DxD

V ″ y = L ″ x − BxB

and

(2.33)

Subtracting both equations gives: y (V ′ − V ″ ) = x (L ′ − L ″ ) + DxD + BxB = x (L ′ − L ″ ) + FxF

(2.34)

which can be rearranged into y=

L″ − L′ V″ − V′

x−

F V″ − V′

xF

(2.35)

Substituting L ″-L ′ and V ″-V ′ from Eq. 2.32 and some rearrangements results in the equation for the feed-line or q-line: y=−

q 1−q

x+

1 1−q

xF

(2.36)

This equation represents a straight line of slope –q/(1 – q) passing through the point (yF , xF) on which every possible intersection point of the two operating lines for a given feed must fall. The position of the line depends only on xF and q′. Thus by changing the reflux ratio the point of intersection is changed but remains on the q-line. From the definition of q′, it follows that the slope of the q-line is governed by the nature of the feed. Examples of the various types of feeds and the slopes of the q-line are illustrated in Fig. 2.16. Note that all the feed lines intersect at one point, which is at y = x = xF . Following the placement of the rectifying section operating line and the q-line, the stripping section operating line is located by drawing a straight line from the point (y = xB , x = xB) on the y = x line through the point of intersection with the q-line and the rectifyingsection operating line as shown in Fig. 2.15.

36

2 Evaporation and distillation

Fig. 2.16 Dependence of q-line on feed condition (for q' see footnote on page 34).

2.4.2 Required number of equilibrium stages 2.4.2.1 Graphical determination of stages and location of feed stage Once the equilibrium line, operating lines and q-line have been plotted, we can continue with the determination of the number of equilibrium stages required for the entire column and the location of the feed stage. The operating lines relate the solute concentration in the vapor passing upward between two stages to the solute concentration in the liquid passing downward between the same two stages. The equilibrium curve relates the solute concentration in the vapor leaving an equilibrium stage to the solute concentration in the liquid leaving the same stage. This makes it possible to determine the required number of stages and the location of the feed stage by constructing a staircase between the operating line and the equilibrium curve, as shown in Fig. 2.17. Although the construction can start either from the top, the bottom or from the feed stage, it is common that the staircase is stepped of from the top and continued all the way to the bottom. Starting from the point (y1 = xD , x0 = xD) on the rectifying section operating line and the y = x line, a horizontal line is drawn to the left until it intersects the equilibrium curve at (y1, x1), the compositions of the equilibrium phases leaving the top equilibrium stage. A vertical line is now dropped until it intersects the operating line at the point (y2, x1), the compositions of the two phases passing each other between stages 1 and 2. The horizontal and vertical line construction is continued down the rectifying section until the feed stage is reached. At the feed stage we switch to the stripping section operating line because of the different operating line in the stripping section due to the

2.4 McCabe-Thiele analysis

37

Fig. 2.17 McCabe-Thiele graphical determination of the number of equilibrium stages.

addition of feed. Although the separation shown in Fig. 2.17 would require exactly 7 equilibrium contacts (6 equilibrium stages plus an equilibrium partial reboiler), in practice hardly ever an integer number of stages result, but rather a fractional stage will appear at the top or at the bottom. Starting the staircase construction from the feed stage to both the top and the bottom stages, at either side generally a fraction of a theoretical stage will be found, see Fig. 2.18a. Either fraction of a theoretical stage can be determined from: fraction =

distance from operating line to xB distance from operating line to equilibrium line

(2.37)

In Fig. 2.17 it is seen that the smallest number of total stages occurs when the transfer is made at the first opportunity after a horizontal line of the staircase crosses the q-line. This is the optimal feed stage location because the separation will require the fewest total number of stages when feed stage 3 is used. For binary distillation the optimum feed stage will always be the stage where the step in the staircase includes the point of intersection of the two operating lines. 2.4.2.2 Limiting conditions For any distillation operation there are infinite combinations of reflux ratios and numbers of theoretical stages possible. The larger the reflux ratio, the fewer theoretical stages are required but the more energy is consumed. For a given combina-

38

2 Evaporation and distillation

Fig. 2.18 Determination of (a) minimum number of theoretical stages and (b) minimum reflux ratio in a McCabe-Thiele diagram.

tion of feed, distillate and bottom compositions, there are two constraints that set the boundary conditions within which the reflux ratio and number of theoretical stages must be. The minimum number of theoretical stages, Nmin , and the minimum reflux ratio, Rmin . As the reflux ratio is increased, the slope of the rectifying section operating line increases to a limiting value of L ′/V ′ = 1, meaning that the system is at total reflux. Correspondingly the slope of the stripping section operating line decreases to a limiting value of L ″/V ″ = 1. The minimum number of theoretical stages occurs when both operating lines coincide with the y = x line and neither the feed composition nor the q-line influences the staircase construction. An example of the McCabe-Thiele construction for this limiting condition is shown in Fig. 2.18a. Because the operating lines are located as far away as possible from the equilibrium curve, a minimum number of stages is required. Column operation with minimum internal gas and liquid flow (i. e. minimum reflux) separates a mixture with the lowest energy input. As the reflux ratio decreases, the intersection of the two operating lines and the q-line moves from the y = x line toward the equilibrium curve. The number of equilibrium stages required increases because the operating lines move closer to the equilibrium curve. Finally a limiting condition is reached when the point of intersection is on the equilibrium curve, as shown in Fig. 2.18b. To reach that stage from either the rectifying section or the stripping section, an infinite number of stages would be required to achieve the desired separation. The corresponding minimum reflux ratio is easily determined from the slope of the limiting operating line (L ′/V ′)min for the rectifying section:

2.4 McCabe-Thiele analysis

Fig. 2.19

39

Typical distillation fixed, operating and total costs as a function of reflux ratio.

L′ Rmin =

L′

(D )min

=

(V ′ )

L′

(V

′

−L

′

)min

min

= 1−

L′

(2.38)

(V ′ )

min

Both of these limits, the minimum number of stages and the minimum reflux ratio serve as valuable guidelines within which the practical distillation conditions must lie. The operating, fixed and total cost of a distillation system are a strong function of the relation of the operating reflux ratio to the minimum reflux ratio. As shown by Fig. 2.19, first the fixed cost decrease by increasing the reflux ration because fewer stages are required but then rises again as the diameter of the column increases at higher vapor and liquid loads. Similarly, the operating cost for energy increase almost linearly as the operating reflux ratio increases. For most commercial operations the optimal operating reflux ratios are in the range of 1.1 to 1.5 times the minimum reflux ratio.

40

2 Evaporation and distillation

2.4.2.3 Fenske and underwood equations Alternative to the graphical determination, approximate values for the minimum number of stages and minimum reflux can also be obtained from the Fenske and Underwood equations, respectively. Both equations are based on the special situation that the relative volatility remains approximately constant throughout the column. The case of the minimum number of theoretical stages, Nmin, is achieved at total reflux, as discussed in the previous section. Total reflux exists when no product is withdrawn: D=B=0

(2.39)

V ′ − L′ = L″ − V ″ = 0

(2.40)

hence

and the slope of both operating lines is unity, see Eqs. 2.24 and 2.26. This means that the composition of the streams passing between two subsequent stages are equal: y2 = x1, y3 = x2

or, generally,

yn + 1 = xn

(2.41)

For a binary equilibrium between component A and the less volatile component B on stage 1 (see Fig. 2.13) Eq. 2.1 applies: y1 = KA ⋅ x1 = KA ⋅ y2

(2.42)

For stage 2 it follows that y1 = KA2 ⋅ y3

(2.43)

Similarly, repeating this procedure for a succession of N theoretical stages the result becomes: y1 = KAN ⋅ yN+1

(2.44)

At the top of the column y1 = x0 = xD for a total condenser while for the total reboiler yN+1 = xN = xB . Hence, the connection between the mole fraction of A and B in the top and bottom product is xD = K AN ⋅ xB

and

1 − xD = K BN ⋅ (1 − xB)

(2.45)

Finally, noting that KA /KB = αAB , Nmin stages are needed to obtain the preset top and bottom compositions: xD 1 − xD

= (αAB) Nmin

xB 1 − xB

(2.46)

Solving the equation for Nmin by logarithms gives the Fenske equation that can be used to estimate the minimum number Nmin of required theoretical stages (Nmin − 1 plates plus the reboiler) in the column:

2.4 McCabe-Thiele analysis

ln

41

xD xB [(1 − xD) / (1 − xB)]

Nmin =

(2.47)

ln αAB

The Fenske equation is also applicable for multi-component mixtures when the relative volatility is based on the light key relative to the heavy key. If the change in relative volatility from the bottom of the column to the top is moderate, a mean value of the relative volatility is generally calculated by taking the average of the relative volatility at the top, αD , and the bottom of the column, αB as follows: αav =

√αB ⋅ αD

(2.48)

In a similar way the minimum reflux ratio Rmin can be obtained analytically from the physical properties of a binary system. Starting again from a material balance (Fig. 2.13), this time over the top portion of the column comprising n theoretical stages, gives: V ′yn + 1 = L ′ xn + DxD

(2.49)

V ′ (1 − yn + 1) = L ′ (1 − xn) + D (1 − xD)

(2.50)

and

Under conditions of minimum reflux, a column has to have an infinite number of plates. For large values of n, approaching the pinch (see Fig. 2.18b), the composition on plate n is equal to that on plate n + 1, so near the pinch yn + 1 = y∞

xn = x∞

and

(2.51)

Application of the equilibrium relation Eq. 2.7, dividing equations (2.49) and (2.50) and introduction of the relative volatility αAB provides: y∞ 1 − y∞

=

αAB x∞

=

1 − x∞

L ′ x∞ + DxD L ′ (1 − x∞) + D (1 − xD)

(2.52)

Solving for L ′/D this rearranges into the Underwood expression for estimating the minimum reflux ratio required to separate the liquid composition on a plate near the pinch into the desired distillate composition:

Rmin =

L′

(D)

min

=

xD 1 − xD − αAB x∞ 1 − x∞ αAB − 1

(2.53)

42

2 Evaporation and distillation

This expression applies to any feed condition q because the relation between the pinch and the feed composition, xF , is not defined yet. For q = 1, saturated liquid, x∞ = xF and the minimum reflux ratio for the desired separation of the feed composition becomes: xD 1 − xD − αAB 1 − xF xF Rmin =

αAB − 1

(2.54)

If the distillate product is required as a pure substance (xD = 1), this equation reduces to: Rmin =

1 (αAB − 1) xF

(2.55)

2.4.2.4 Use of murphree efficiency The McCabe-Thiele method assumes that the two phases leaving each stage are in thermodynamic equilibrium. In industrial equipment the equilibrium is usually not fully approached. The most commonly used stage efficiencies used to describe individual tray performance for individual components are the Murphree vapor

Fig. 2.20 Application of the Murphree plate efficiency in the McCabe-Thiele construction.

2.4 McCabe-Thiele analysis

43

and liquid efficiencies. The Murphree vapor efficiency EMV for stage n is defined as the ratio of the actual change in vapor composition over the change in vapor composition for an equilibrium stage: EMV =

yn − yn + 1

(2.56)

yn* − yn + 1 where y*n is the composition of the hypothetical vapor phase in equilibrium with the liquid composition leaving stage n. Once the Murphree efficiency is known for every stage, it can easily be used on a McCabe-Thiele diagram. The denominator represents the vertical distance from the operating line to the equilibrium line while the numerator is the vertical distance from the operating line to the actual outlet concentration. In stepping of stage, the Murphree vapor efficiency represents the fraction of the total vertical distance to move from the operating line to the equilibrium line. As only EMV of the total vertical path is traveled, we get the result shown in Fig. 2.20. 2.4.3 Energy requirements Following the determination of the feed condition, reflux ratio and number of theoretical stages estimates of the heat duties of the condenser and reboiler can be made. When the column is well insulated, all of the heat transfer takes place in the condenser and reboiler, column pressure is constant, the feed is at the bubble point and a total condenser is used, the energy balance over the entire column gives: Reboiler In = Condenser Out

(2.57)

Q R = QC

(2.58)

or

The energy balance can by approximated by applying the assumptions of the McCabe-Thiele method, yielding for the reboiler and condenser duty: Q R = QC = D (R + 1) ΔHvap

(2.59)

If saturated steam is the heating medium for the reboiler, the required steam rate becomes: Φm, steam = where

ΔHvap, steam Msteam

Q R ⋅ Msteam ΔHvap, steam

(2.60)

is the specific enthalpy of vaporization of water (≈ 2100 kJ/kg).

44

2 Evaporation and distillation

The cooling water rate for the condenser is: Φm,C =

QC CP, water (Tout − Tin)

(2.61)

where CP,water is the specific heat capacity of water (≈ 4.2 kJ/kg K). In general the cost of cooling water can be neglected during a first evaluation because the annual cost of reboiler steam are an order of magnitude higher.

2.5 Advanced distillation techniques 2.5.1 Batch distillation In most large chemical plants distillations are run continuously. For small production units where most chemical processes are carried out in batches it is more convenient to distil each batch separately. In batch distillation, a liquid mixture is charged to a vessel where it is heated to the boiling point. When boiling begins the vapor is passed through a fractionation column and condensed to obtain a distillate product, as indicated in Fig. 2.21. As with continuous distillation the

Fig. 2.21 Schematic of a batch distillation unit.

2.5 Advanced distillation techniques

45

purity of the top product depends on the still composition, the number of plates of the column and on the reflux ratio used. In contrast to continuous distillation, batch distillation is usually operated with a variable amount of reflux. Because the lower-boiling components concentrate in the vapor and the remaining liquid gradually becomes richer in the heavier components, the purity of the top product will steadily drop. This is generally compensated by a gradual increase in reflux ratio during the distillation process to maintain a constant quality of the top product. To obtain the maximum recovery of a valuable component, the charge remaining in the still after the first distillation may be added to the next batch. The main advantage of batch distillation is that multiple liquid mixtures can be processed in a single unit. Different product requirements are easily taken into account by changing the reflux ratio. Even multicomponent mixtures can be separated into the different components by a single column when the fractions are collected separately. An additional advantage is that batch distillation can also handle sludge and solids. The main disadvantages of batch distillation are that for a given product rate the equipment is larger and the mixture is exposed to higher temperature for a longer time. This increases the risk of thermal degradation or decomposition. Furthermore, it requires more operator attention, energy requirements are higher and its dynamic nature makes it more difficult to control and model. 2.5.2 Continuous separation of multiple product mixtures Very often a separation unit incorporated in the process separates the stream into several products by means of several columns. Where the columns are located in the flow diagram will have a very great influence on the economics of the separation unit. Case studies have shown that operating costs alone may vary by a factor of two depending on the flow sheet chosen. If each component of a multicomponent distillation is to be essentially pure when recovered, the number of required columns is equal to the number of components minus one. Thus, a three-component mixture requires two and a four-component mixture requires three separate columns. Those columns can be arranged in many different ways as illustrated in Fig. 2.22 for the two possible separation configurations of a ternary mixture. Note that it takes a sequence of two ordinary distillation columns to separate a mixture in three products. Because of the many possible flow sheet configurations for mixtures with more than 3 components it would be far too complex to find the best flow diagram by systematic study. Fortunately some heuristic rules are available to provide guidelines for defining flow diagrams to accelerate the search for the optimal separation sequence of multi-component mixtures: 1. Remove corrosive and hazardous materials first 2. First eliminate majority components 3. Start out with easy separations

46

2 Evaporation and distillation

Fig. 2.22 Possible distillation configurations for separating ternary mixtures.

4. Sequences which remove the components one by one in column overheads should be favored 5. Give priority to separations which give a more nearly equimolal division of the feed between the distillate and bottoms product 6. End up with difficult separations such as azeotropic mixtures 7. Separations involving very strict product specifications should be reserved until late in a sequence 8. Favor sequences which yield the minimum necessary number of products 9. Avoid vacuum distillation and refrigeration if possible

2.5.3 Separation of azeotropes When, due to minimal difference in boiling point and/or highly non-ideal liquid behavior the relative volatility becomes lower than 1.1, ordinary distillation may be uneconomic and in case an azeotrope forms even impossible. In that event, enhanced distillation techniques should be explored. For these circumstances the most often used technique is extractive distillation where a large amount of a relatively high-boiling solvent is added to increase the relative volatility of the key components in the feed mixture. In order to maintain a high concentration throughout the column, the solvent is generally introduced above the feed entry and a few trays below the top. It leaves the bottom of the column with the less volatile product and is recovered in a second distillation column as shown in Fig. 2.23. Because the high-boiling solvent is easily recovered by distillation when selected in such a way that no new azeotropes are formed, extractive distillation is less complex and more widely used than azeotropic distillation. In homogeneous azeotropic distillation an entrainer is added to the mixture that forms a homogeneous minimum- or maximum-boiling azeotrope with one or

2.5 Advanced distillation techniques

47

Fig. 2.23 Extractive distillation separation of toluene from methylcyclohexane (MCH) using high boiling polar solvents such as NMP, sulfolane, phenol, glycols.

more feed components. The entrainer can be added everywhere in the column. If an entrainer is added to form an azeotrope, the azeotrope will exit the column as the overhead or bottom product leaving behind component(s) which may be recovered in the pure state. A classical example of azeotropic distillation is the recovery of anhydrous ethanol from aqueous solutions. Organic solvents such as benzene or cyclohexane are used to form desirable azeotropes, allowing the separation to be made. As shown in Fig. 2.24, the ethanol azeotrope from the crude column overhead is fed to the azeotropic distillation column where cyclohexane is used to form an azeotrope with water. Ethanol is recovered as the bottom product. The overhead azeotropic cyclohexane/water mixture is condensed where water-rich and cyclohexane-rich liquid phases are formed. Residual cyclohexane is removed from the water rich phase in a stripper. It is well known that many minimum-boiling azeotrope-containing mixtures allow the position of the azeotrope to be shifted a change in system pressure. This effect can be exploited to separate a binary azeotrope containing mixture when appreciably changes (> 5 mol%) in azeotropic composition can be achieved over a moderate pressure range. Pressure swing distillation uses a sequence of two columns operated at different pressures for the separation of pressure-sensitive azeotropes. This is illustrated in Fig. 2.25 where the effect of pressure on the

48

2 Evaporation and distillation

Fig. 2.24 Dehydration of alcohol by azeotropic distillation with solvents such as benzene or cyclohexane as entrainer.

Fig. 2.25 (a) Pressure-swing distillation: T-y-x curves for minimum-boiling azeotrope at pressures P1 and P2 and (b) distillation sequence for minimum-boiling azeotrope.

2.5 Advanced distillation techniques

49

temperature and composition of a minimum boiling azeotrope is given. The binary azeotrope can be crossed by first separating the component boiling higher than the azeotrope at low pressure. The composition of the overhead should be as close as possible to that of the azeotrope at this pressure. As the pressure is increased the azeotropic composition moves toward a higher percentage of A and component B can be separated from the azeotrope as the bottom product in the second column. The overhead of the second column is returned to the first column.

Nomenclature B CP D DAB EMV F ΔHvap Ki L NOV N p P0 Ptot q Q R T x, y, z V Φm αij γ

bottom product flow specific heat capacity top product flow coefficient of diffusion of A in a mixture of A and B Murphree or plate efficiency, Eq. 2.56 feed flow molar heat of vaporization distribution coefficient of component i liquid flow (in rectifying section L ′, in stripping section L ″ ) overall number of gas-phase transfer units number of (theoretical) stages partial pressure saturation pressure total pressure fraction saturated liquid feed, Eq. 2.13 heat flow reflux ratio L ′/D, Eq. 2.23 temperature mole fraction (liquid, vapor, feed) vapor flow (in rectifying section V ′, in stripping section V ″) mass flow rate selectivity, relative volatility or equilibrium constant, Eqs. 1.5, 2.2, 2.7 , 2.8 activity coefficient

Indices A, B, i, j B C D F L n, m R V

components bottom condenser distillate feed liquid stage number reboiler vapor

[mol s–1] [J kg–1K–1] [mol s–1] [m2s–1] [–] [mol s-1] [J mol–1] [–] [mol s-1] [–] [–] [N m–2] [N m–2] [N m–2] [–] [J s-1] [–] [K] [–] [mol s–1] [kg s–1] [–] [–]

50

2 Evaporation and distillation

Exercises 1. In Vapor-Liquid Equilibrium Data Collection the following form of Antoine-equation is used log

Pi0

B′

′

=A −

T + C′

with Pi0 = saturation pressure in mmHg and temperature T in °C (760 mmHg = 1 atm) For benzene in benzene-toluene mixtures the following values are reported: A ′ = 6.87987 [–], B ′ = 1196.760 [–]

and

C ′ = 219.161 [ °C].

Calculate the constants A, B and C in the Antoine-equation as defined in Eq. 2.9: ln Pi0 = A −

B T+C

with pressure units in atm and temperature in °C. 2. With temperature in degrees Fahrenheit and pressure in pounds per sq. inch, the following Antoine constants apply for benzene A ′ = 5.1606, B ′ = 2154.2, C ′ = 362.49 (psi, °F,

10

log P expression)

Noting that 1 atm = 14.696 psi = 1.01325 bar and T (°F) = T ( °C) · 9/5 + 32, calculate the saturation pressure of benzene in bar at 80.1 °C. 3. Vapor-liquid equilibrium data for benzene-toluene are given at 1 atm and at 1.5 atm in the T-x diagram on page 51 (Fig. 2.26). Expressing the saturated vapor pressure in atm as a function of temperature in K, the Antoine constants for benzene and toluene are, respectively A = 9.2082,

B = 2755.64,

C = –54.00

A = 9.3716,

B = 3090.78,

C = –53.97 (atm, K, Eq. 2.9).

and

Check the phase compositions at 100 °C and total pressures of 1.0 and 1.5 atm. 4. A liquid benzene-toluene mixture with z = 0.40 should produce a liquid with x = 0.35 in a flash drum at 1.0 bar. a. Calculate the required feed temperature b. Calculate the equilibrium vapor-liquid ratio in the flash drum. 5. We wish to flash distill isothermally a mixture containing 45 mole% of benzene and 55 mole% of toluene. Feed rate to the still is 700 moles/h. Equilibrium data for the benzene-toluene system can be approximated with a constant relative volatility of 2.5, where benzene is the more volatile component. Operation of the still is at 1 atm. a. Plot the y-x diagram for benzene-toluene. b. If 60 % of the feed is evaporated, find the liquid and vapor compositions c. If we desire a vapor composition of 60 mole%, what is the corresponding liquid composition and what are the liquid and vapor flow rates? d. Find the compositions and flow rates of all unknown streams for a two stage flash cascade where 40 % of the feed is flashed in the first stage and the liquid product is sent to a second flash chamber where 30 % is flashed.

Exercises

51

Fig. 2.26

6. Distillation is used to separate pentane from hexane. The feed amounts 100 mol/s and has a mole ratio pentane/hexane = 0.5. The bottom and top products have the compositions xB = 0.05 and xD = 0.98. The reflux ratio is 2.25. The column pressure is 1 bar. The feed, at the bubbling point, enters the column exactly on the feed tray. The tray temperature is equal to the feed temperature. The pentane vapor pressure is given by: (bar) p 50 = exp 11 ⋅ 1 − 310 ( ( T (K)))

(1 atm = 1.013 bar)

The vapor pressure of hexane is 1/ 3 of the pentane vapor pressure over the whole temperature range. The average density of liquid pentane and hexane amount 8170 resp. 7280 mol/m3. The heat of vaporization amounts 30 kJ/mol. The distance between the trays amounts 0.50 m. a. Calculate the feed temperature b. Calculate the vapor stream from the reboiler c. Calculate the required energy in the reboiler d. Construct the y-x diagram e. Construct the operating lines and locate the feed line f. Determine the number of equilibrium stages g. Determine the height of the column 7. Methanol (M) is to be separated from water (W) by atmospheric distillation. The feed contains 14.46 kg/h methanol and 10.44 kg/h water. The distillate is 99 mol% pure while the bottom product contains 5 mol % of water. The feed is subcooled such that q = 1.12. a. Determine the minimum number of stages and minimum reflux

52

2 Evaporation and distillation

b. Determine the feed stage location and number of theoretical stages required for a reflux ratio of R = 1. Vapor-liquid equilibrium data (1 atm, mole fraction methanol) x 0.0321 0.0523 0.075 0.154 0.225 0.349 0.813 0.918 y 0.1900 0.2940 0.352 0.516 0.593 0.703 0.918 0.963 8. A feed to a distillation unit consists of 50 mol% benzene in toluene. It is introduced to the column at its bubble point to the optimal plate. The column is to produce a distillate containing 95 mol% benzene and a bottoms of 95 mol% toluene. For an operating pressure of 1 atm, calculate: a. the minimum reflux ratio b. the minimum number of equilibrium stages to carry out the desired separation c. the number of actual stages needed, using a reflux ratio (L ′/D) of 50 % more than the minimum, c. the product and residue stream in kilograms per hour of product if the feed is 907.3 kg/h e. the saturated steam required in kilograms per hour for heat to the reboiler using enthalpy data below Steam: ΔHvap = 2000 kJ/kg Benzene: ΔHvap = 380 kJ/kg Toluene: ΔHvap = 400 kJ/kg Vapor-liquid equilibrium data (1 atm, mole fraction benzene) x 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 y 0.21 0.37 0.51 0.64 0.72 0.79 0.86 0.91 0.96

3 Absorption and stripping

3.1 Introduction In absorption (also called gas absorption, gas scrubbing and gas washing) a gas mixture is contacted with a liquid (the absorbent or solvent) to selectively dissolve one or more components by transfer from the gas to the liquid. The components transferred to the liquid are referred to as solutes or absorbates. The operation of absorption can be categorized on the basis of the nature of the interaction between absorbent and absorbate into the following three general types: 1. Physical solution. In this case, the component being absorbed is more soluble in the liquid absorbent than the other gases with which it is mixed but does not react chemically with the absorbent. As a result, the equilibrium concentration in the liquid phase is primarily a function of partial pressure in the gas phase and temperature. Examples are the drying of natural gas with diethylene glycol or the recovery of ethylene oxide and acrylonitrile with water from the reactor product stream. 2. Reversible reaction. This type of absorption is characterized by the occurrence of a chemical reaction between the gaseous component being absorbed and a component in the liquid phase to form a compound that exerts a significant vapor pressure of the absorbed component. The most important industrial example is the removal of acid gases (CO2, H2S) with mono- or diethanolamine solutions. 3. Irreversible reaction. In this case, a reaction occurs between the component being absorbed and a component in the liquid phase, which is essentially irreversible. Sulfuric acid and nitric acid production by SO3 and NO2 absorption in water is the most widely used example of this application. The use of physical absorption processes is usually preferred whenever feed gases are present in large amounts at high pressure and the amount of the component to be absorbed is relatively large. Chemical absorption usually has a much more favorable equilibrium relationship than physical absorption and is therefore often preferred when the components to be separated from feed gases are present in small concentrations and at low partial pressures Gas absorption is usually carried out in vertical countercurrent columns as shown in Fig. 3.1. The solvent is fed at the top whereas the gas mixture enters from the bottom. The absorbed substance is washed out by the lean solvent and leaves the absorber at the bottom as a liquid solution. Usually, the absorption

54

3 Absorption and stripping

Fig. 3.1 Basic scheme of an absorption installation with stripping for regeneration.

column operates at a pressure higher than atmospheric, taking advantage of the fact that gas solubility increases with pressure. The loaded solvent is often recovered in a subsequent stripping or desorption operation. After preheating, the rich solvent is transported to the top of a desorption column that usually operates under lower pressure than in the absorption column. This second step is essentially the reverse of absorption in which the absorbate is removed from the solvent. Desorption can be achieved through a combination of methods: 1. flashing the solvent to lower the partial pressure of the dissolved components 2. reboiling the solvent, to generate stripping vapor by evaporation of part of the solvent 3. stripping with an inert gas or steam Wide use is made of desorption by pressure reduction because the energy requirement are low. After depressurization, stripping with an inert gas is often more economic than thermal regeneration. However, because stripping is not perfect, the absorbent recycled to the absorber contains residual amounts of the absorbed solute. The desired purity of the absorbent determines the final costs of desorption. The necessary difference between the partial pressure of the absorbed key component over the regenerated solution and the purified gas serves as a criterion for determining the dimensions of the absorption and desorption equipment. The lean solvent, devoid of gas, flows through the heat exchanger, where part of the

3.2 The aim of absorption

55

heat needed for heating the rich solvent is recovered, and then through a second heat exchanger, where it is cooled down to a desired temperature and flows into the absorption column. Usually a small amount of fresh solvent should be added to the column to replenish the solvent which was partly evaporated in the desorption column or underwent irreversible chemical reactions which take place in the whole system.

3.2 The aim of absorption Generally the commercial purpose of absorption processes can be divided into gas purification or product recovery, depending on whether the absorbed or the unabsorbed portion of the feed gas has the greater value. Typical gas purification applications are listed in Table 3.1. The removal of CO2 from hydrogen gas in ammonia production and the removal of acid gases (CO2, H2S) from natural gas are some of the most widespread applications that are being improved continuously by the development of new solvents, process configurations and design techniques. In both applications stripping is used for absorbent regeneration. Examples of absorption processes for product recovery are listed in Table 3.2. The absorption of SO3 and NOx in water to make concentrated sulfuric respectively nitric acid are probably the most widely used product recovery applications of absorption. Other frequently encountered examples are the recovery of various products from a gaseous product stream by inert absorbents such as water. In some cases the absorber is used as a reactor where the desired chemical compound is obtained by a liquid phase reaction of the absorbed gases. An illustration of such a process is the production of urea from CO2 and ammonia.

Table 3.1

Typical applications of absorption for gas purification.

Impurity

Process

Absorbent

Ammonia Carbon Dioxide and Hydrogen Sulfide

Indirect Process (Coke Oven Gas) Ethanolamine

Water Mono- or Diethanolamine in Water Potassium Carbonate and Activator in Water Polyethylene Glycol Dimethyl Ether Cuprous Ammonium Carbonate and Formate in Water Water Toluene Cyclohexane

Benfield Selexol Carbon Monoxide

Copper Ammonium Salt

Hydrogen Chloride Toluene Cyclohexane

Water Wash Toluene Scrubber Scrubber

56

3 Absorption and stripping

Table 3.2

Product

Typical applications of absorption for product recovery.

Process

Acetylene

Steam Cracking of Hydrocarbons (Naphtha) Acrylonitrile Ammoxidation of Propylene Maleic Anhydride Butane Oxidation Melamine Urea Decomposition Nitric Acid Ammonia Oxidation (NOx Absorption) Sulfuric Acid Contact Process (SO3 Absorption) Urea Synthesis (CO2 and NH3 Absorption)

Absorbent Dimethylformamide Water Water Water Water Water Ammonium Carbamate Solution

3.3 General design approach Both absorption and stripping can be operated as equilibrium stage operations with contact of liquid and vapor. The plate towers can be designed by following an adaptation of the McCabe-Thiele method. Packed towers can be designed by the use of HETP or preferably by mass transfer considerations. In both absorption and stripping a separate phase is added as the separating agent. Thus the columns are simpler than those in distillation in that reboilers and condensers are normally not used. Design or analysis of an absorber (or stripper) requires consideration of a number of factors, including: 1. 2. 3. 4. 5.

6. 7. 8. 9. 10.

Entering gas (liquid) flow rate, composition, temperature, and pressure Desired degree of recovery of one or more solutes Choice of absorbent (stripping agent) Operating pressure and temperature, and allowable gas pressure drop Minimum absorbent (stripping agent) flow rate and actual absorbent (stripping agent) flow rate as a multiple of the minimum rate needed to make the separation Number of equilibrium stages Heat effects and need for cooling (heating) Type of absorber (stripper) equipment Height of absorber (stripper) Diameter of absorber (stripper)

The initial step in the design of the absorption system is selection of the absorbent and the overall process to be employed. There is no simple analytical method for accomplishing this step. In most cases, more than one solvent can meet the process requirements, and the only satisfactory approach is an economic evaluation, which may involve the complete but preliminary design and cost estimate for more than one alternative. The ideal absorbent should:

3.4 Absorption and stripping equilibria

57

• have a high solubility for the solute(s) to minimize the need for absorbent • have a low volatility to reduce the loss of absorbent and facilitate separation of solute(s) • be stable to maximize absorbent life and reduce absorbent makeup requirement • be noncorrosive to permit use of common materials of construction • have a low viscosity to provide low pressure drop and high mass and heat transfer rates • be nonfoaming when contacted with the gas • be nontoxic and nonflammable to facilitate its safe use • be available, if possible within the process, or be inexpensive. The most widely used absorbents are water, hydrocarbon oils, and aqueous solutions of acids and bases. For stripping the most common agents are water vapor, air, inert gases, and hydrocarbon gases. Once an absorbent is selected, the design of the absorber requires the determination of basic physical property data such as density, viscosity, surface tension and heat capacity. The fundamental physical principles underlying the process of gas absorption are the solubility and heat of solution of the absorbed gas and the rate of mass transfer. Information on both must be available when sizing equipment for a given application. In addition to the fundamental design concepts based on solubility and mass transfer, many practical details have to be considered during actual plant design. The second step is the selection of the operating conditions and the type of contactor. In general, operating pressure should be high and temperature low for an absorber, to minimize stage requirements and/or absorbent flow rate. Operating pressure should be low and temperature high for a stripper to minimize stage requirements or stripping agent flow rate. However, because maintenance of a vacuum is expensive, strippers are commonly operated at a pressure just above ambient. A high temperature can be used, but it should not be so high as to cause undesirable chemical reactions. Choice of the contactor may be done on the basis of system requirements and experience factors such as those discussed in paragraph 3.6. Following these decisions it is necessary to perform material and heat balance calculations around the contactor, define the mass transfer requirements, determine the height of packing or number of trays and calculate contactor size to accommodate the liquid and gas flow rates with the selected column internals.

3.4 Absorption and stripping equilibria 3.4.1 Gaseous solute solubilities The most important physical property data required for the design of absorbers and strippers are gas-liquid equilibria. Since equilibrium represents the limiting condition for any gas-liquid contact, such data are needed to define the maximum gas purity and rich solution concentration attainable in absorbers, and the maxi-

58

3 Absorption and stripping

mum lean solution purity attainable in strippers. Equilibrium data are also needed to establish the mass transfer driving force, which can be defined simply as the difference between the actual and equilibrium conditions at any point in a contactor. At equilibrium, a component of a gas in contact with a liquid has identical fugacities in both the gas and liquid phase. For ideal solutions Raoultʼs law applies, see also Eq. 2.6: yA =

PA0 Ptot

xA

(3.1)

where yA is the mole fraction of A in the gas phase, Ptot is the total pressure, PA0 is the saturation pressure, the vapor pressure of pure A, and xA is the mole fraction of A in the liquid. For non-ideal mixtures Raoultʼs law modifies into: yA =

γA∞ PA0 Ptot

xA

(3.2)

where γA∞ is the activity coefficient of solute A in the absorbent at infinite dilution. A more general way of expressing solubilities is through the dimensionless vaporliquid distribution coefficient K, defined in the previous chapter (Eq. 2.1) yA ≡ KA xA

(3.3)

Values of distribution coefficients are widely employed to represent hydrocarbon vapor-liquid equilibria in absorption and distillation calculations. Correlations for and experimental data on the distribution coefficients of hydrocarbons are available from various sources. For moderately soluble gases with relatively little interaction between the gas and liquid molecules equilibrium data are usually represented by Henryʼs law: pA = Ptot yA = HA xA

(3.4)

where pA is the partial pressure of A in the gas phase. HA is a Henry constant1, which has the units of pressure. The Henryʼs law constants depend upon temperature and usually follow an Arrhenius relationship: HA = HA0 exp

−ΔHvap

( RT )

(3.5)

A plot of ln HA versus 1/T will then give a straight line. Usually a Henry constant increases with temperature, but is relatively independent of pressure at moderate levels. In general, for moderate temperatures, gas solubilities decrease with an

1

Note that we use HA = Henry coefficient of a component A, whereas ΔHvap = HV – HL refers to enthalpies.

3.4 Absorption and stripping equilibria

59

Fig. 3.2 Solubilitles of various gases in water expressed as the reciprocal of the Henry’s Law constant (adapted from [15]).

increase in temperature. Henryʼs constants for many gases and solvents are tabulated in various literature sources. Examples of Henryʼs constants for a number of gases in pure water are given in Fig. 3.2. 3.4.2 Minimum absorbent flow For each feed gas flow rate, absorbent composition, extent of solute absorption, operating pressure and operating temperature, a minimum absorbent flow rate exists that corresponds to an infinite number of countercurrent equilibrium contacts between the gas and liquid phases. As a result a tradeoff exists in every design problem between the number of equilibrium stages and the absorbent flow rates at rates greater than the minimum value. This minimum absorbent flow rate Lmin is obtained from a mass balance over the whole absorber, assuming equilibrium is obtained between in incoming gas and outgoing absorbent liquid in the bottom of the column. An overall mass balance over the column illustrated by Fig. 3.3 gives the following result: Gyin + Lxin = Gyout + Lxout

(3.6)2

The lower the absorbent flow, the higher xout . The theoretically highest possible concentration in the liquid xout determines the minimum absorbent flow, such as an infinitely long column where equilibrium can be assumed between incoming gas and outgoing liquid (xout = yin /K): 2

Note that in absorption the gas phase ususally is denoted by G instead of V(apor) as in the previous chapter.

60

3 Absorption and stripping

Fig. 3.3 Continuous, steady-state operation in a counter-current column.

L min = G ⋅

yin − yout xmax − xin

= G⋅

yin − yout yin − xin K

(3.7)

A similar derivation of the minimum stripping gas flow rate Gmin for a stripper results in an analogous expression: Gmin = L ⋅

xin − xout ymax − yin

= L⋅

xin − xout Kxin − yin

(3.8)

3.5 Absorber and stripper design 3.5.1 Operating lines for absorption The McCabe-Thiele diagram is most useful when the operating line is straight. This requires that the energy balances are automatically satisfied and the ratio of liquid to vapor flow rate is constant. In order to have the energy balances automat-

3.5 Absorber and stripper design

61

ically satisfied, we must assume that the heat of absorption is neglible and the operation is isothermal. These two assumptions will guarantee satisfaction of the enthalpy balances. When the gas and liquid stream are both fairly dilute (say < 5 %), the assumptions will probably be satisfied. We also desire a straight operating line at higher concentrations. This will automatically be true if we assume that the solvent is nonvolatile, the carrier gas is insoluble and define liquid and gas streams in terms of solute free solvent and carrier gas: L′ G′

=

moles nonvolatile solvent /s

(3.9)

moles insoluble carrier gas /s

The results of these last two assumptions are that the mass balances for the solvent and carrier gas become: LN′ = L n′ = L 0′ = L ′ = constant and

(3.10) GN′ + 1 = Gn′ = G1′ = G ′ = constant

Now overall flow rates of gas and liquid are not used because in more concentrated mixtures a significant amount of solute may be absorbed, which would change gas and liquid flow rates. This would result in a curved operating line. Using L′ (= moles of nonvolatile solvent/s) and G′ (= moles insoluble carrier gas/ s), we must define compositions in such a way that we can write a mass balance for solute B. The correct way to do this is to define the compositions as mole ratios: Y=

moles solute B in gas

and

moles pure carrier gas

X=

moles solute B in liquid moles pure solvent S

(3.11)

The mole ratios Y and X are related to the usual mole fractions by: Y=

y 1−y

and

X=

x 1−x

(3.12)

Substitution of X and Y into the ideal equilibrium expression, Eq. 2.7, gives the equilibrium line expressed in mole ratios: Y = α⋅X

(3.13)

which represents a straight line through the origin. Note that both Y and X can be greater than unity. With mole ratio units we obtain for the gas and liquid stream leaving stage n:

62

3 Absorption and stripping

Fig. 3.4 Top section mass balance for an absorber.

Yn G ′ = =

moles B in gas stream moles carrier gas ⋅ moles carrier gas s moles B in gas stream

(3.14)

s

and Xn L ′ = =

moles B liquid stream moles solvent ⋅ moles solvent s moles B liquid stream

(3.15)

s

Thus we can easily write the steady-state mass balance, moles solute B in s–1 = moles solute B out s–1, in these units. The mass balance around the top of the column using the mass balance envelope shown in Fig. 3.4 is: Yn + 1 G ′ + X 0 L ′ = X n L ′ + Y1 G ′

(3.16)

Solving for Yn+1 we obtain: Yn + 1 =

L′ G′

Xn + Y1 −

(

L′ G′

X0

)

(3.17)

3.5 Absorber and stripper design

63

Fig. 3.5 McCabe-Thiele diagram for absorption.

This is a straight line with slope L ′ /G ′ and intercept Y1 – (L ′ /G ′) · X0 . It is the operating line for absorption. Thus if we plot ratios Y vs X we have a McCabeThiele type of graph as shown in Fig. 3.5. The steps in the procedure are: 1. Plot Y vs X equilibrium data (convert from fractions to ratioʼs) 2. Values of X0 , YN+1, Y1 and L ′ /G ′ are known. Point (X0 , Y1) is on the operating line since it represents passing streams; Xn follows from an overall balance. 3. Slope is L ′ /G ′. Plot operating line. 4. Starting at stage 1, step off stages: equilibrium, operating line, equilibrium, etc. Note that the operating line in absorption is above the equilibrium line, because solute is transferred from the gas to the liquid. In distillation we had material (the more volatile component) transferred from liquid to gas, and the operating line was below the equilibrium curve. The Y = X line has no significance in absorption. As usual the stages are counted at the equilibrium curve. The minimum L ′ /G ′ ratio corresponds to a value of XN leaving the bottom of the tower in equilibrium with YN+1, the solute concentration in the feed gas. It takes an infinite number of stages for this equilibrium to be achieved. This minimal L ′ /G ′ ratio can be derived from the McCabe-Thiele diagram as shown in Fig. 3.6. The selection of the actual operating absorbent flow rate is based on some multiple of L ′min , typically between 1.1 and 2. A value of 1.5 is usually close to the economically optimal conditions.

64

3 Absorption and stripping

Fig. 3.6 Determination of the minimum (L / G) ratio for absorption from a McCabe-Thiele diagram.

Fig. 3.7 McCabe-Thiele diagram for stripping.

3.5.2 Stripping analysis The number of equilibrium stages for the stripper is determined in a manner similar to that for absorption. Since stripping is very similar to absorption we expect a similar result. The mass balance for the column shown in Fig. 3.7 is the same as for absorption and the operating line remains:

3.5 Absorber and stripper design

Yn + 1 =

L′ G′

Xn +

Y1 −

(

L′ G′

X0

)

65

(3.18)

For stripping we know X0 , XN , YN+1 and L ′ /G ′, Y1 follows from an overall balance. Since (XN , YN+1) is a point on the operating line, we can plot the operating line and step off stages. This is illustrated in Fig. 3.7. Note that the operating line is below the equilibrium curve because solute is transferred from liquid to gas. This is therefore similar to the stripping section of a distillation column. A maximum L ′ /G ′ ratio can be defined. This corresponds to the minimum amount of stripping gas. Start from the known point (YN+1, XN) and draw a line to the intersection of X = X0 and the equilibrium curve. For a stripper, Y1 > YN+1 while the reverse is true in absorption. Thus the top of the column is on the right side in Fig. 3.7 but on the left side in Fig. 3.5. 3.5.3 Analytical Kremser solution McGabe-Thiele diagrams can be used to find graphically the number of equilibrium stages, no matter the operating and equilibrium expressions are linear or not. In case of linear and equilibrium lines, an analytical solution exists. When the solution is quite dilute (less than 1 % in both gas and liquid), the total liquid and gas flow rates will not change significantly since only a small amount of solute is transferred. Now the entire analysis can be done with mole (or mass) fractions and constant total flow rates G and L. Alternatively, when the use of mole ratioʼs and pure carrier gas and solvent flows result in linear operating and equilibrium lines, the following analysis applies equally well3. In this case the column shown in Fig. 3.8 will look like the one in Fig. 3.4 with G ′ = G and L ′ = L. The mass balance around any stage n in terms of mole fractions and total flows: yn + 1 G + xn − 1 L = xn L + yn G

(3.19)

which can be rewritten as: L G

=

yn + 1 − yn xn − xn − 1

(3.20)

is essentially the same as Eq. 3.16 except that the units are different. In case of the straight equilibrium line, i.e. at low values of xn and yn, the equilibrium equation Eq. 2.7 is reduced to: After conversion of mole fractions x, y into mole ratioʼs X, Y and using L ′, G ′ instead of total flows L, G.

3

66

3 Absorption and stripping

Fig. 3.8 Coding of equilibrium stages, streams and fractions for the derivation of the Kremser equations.

(3.21)4

yn = K xn

Eqs. 3.20 and 3.21 can be combined and, introducing the absorption factor A, rewritten as: A≡

L KG

=

yn + 1 − yn K (xn − xn − 1)

=

yn + 1 − K xn yn − K xn − 1

(3.22)

Note that this absorption factor A is different from the separation factor as defined in Eq. 1.1. For stage n − 1 we find in an analogous way: yn − K xn − 1 yn − 1 − K xn − 2

=

L KG

=A

(3.23)

Elimination of xn–1 from Eqs. 3.22 and 3.23 gives: yn + 1 − K xn yn − 1 − K xn − 2

= A2

(3.24)

Continuing in a similar way we find for a cascade with N contacting equilibrium stages: yN + 1 − K xN y1 − K x0 4

=

yin − K xout yout − K xin

= AN

Here the equilibrium constant K replaces α = KA / KB in Eq. 2.8.

(3.25)

3.5 Absorber and stripper design

67

Using the overall mass balance xout can be eliminated from Eq. 3.25: xout = xin +

G

yin −

L

G L

yout

(3.26)

resulting in: AN =

yin − K xin

1−

[(yout − K xin)(

1 A)

+

1

(3.27)

A]

For specified values of yin , yout and xin Eq. 3.27 is rewritten to calculate the number of equilibrium stages Nts : ln

yin − K xin 1 1 1− + [yout − K xin ( A) A ]

Nts =

(3.28)

ln A

This equation is known as the Kremser-equation for absorption. Eq. 3.27 also allows calculating the fraction of the feed that is absorbed in a column with N equilibrium stages. Noting that yin – yout is the actual change in gas composition and yin – K · xin is the maximum possible change in composition, i.e. if gas leaving is in equilibrium with entering liquid, then from Eq. 3.27 the fraction absorbed fA can be derived: Fraction of a solute absorbed = fA =

yin − yout yin − K xin

=

AN + 1 − A AN + 1 − 1

(3.29)

Following an analogous derivation for stripping of a liquid by a gas, we are now interested in the change in liquid phase concentrations. Eq. 3.25 is rewritten in terms of mole fractions solute in liquid: xin −

yout K =

xout

yin − K

1 A

N

= SN

(3.30)

where the solute stripping factor S is defined by S ≡ with the overall mass balance provides now: yout = yin +

L G

x in −

L G

x out

K⋅G L

. Elimination of yout

(3.31)

68

3 Absorption and stripping

resulting in: xin − S Nts =

[ (x

out

yin K

yin − K

)

1

1−

(

S)

+

1 S

(3.32)

]

For specified values of xin , xout and yin, the required number of equilibrium stages Nts is then calculated from the Kremser-equation for stripping: xin − ln

Nts =

[ (x

out

−

yin K yin K

)

1−

(

1 1 + S) S

] (3.33)

ln S

Analogously to the fraction absorbed, the fraction of solute stripped, fS, is defined as the ration of the actual change in liquid composition and the maximum possible change. The degree of stripping obtained in a stripper with Nts equilibrium stages: Fraction of a solute stripped = fS =

xin − xout yin xin − K

=

S Nts + 1 − S S Nts + 1 − 1

(3.34)

Values of L and G in moles per unit time may be taken as the entering values. Values of K depend mainly on temperature, pressure and liquid phase composition.

3.6 Industrial absorbers The main purpose of various industrial absorbers is to ensure large gas-liquid mass transfer area and to create such conditions that a high intensity of mass transfer is achieved. Although small-scale processes sometimes use batch-wise operation where the liquid is placed in the equipment and only gas is flowing, the continuous method of absorption is usually applied in large-scale industrial processes. There are various criteria for classifying absorbers. It seems that the best one is a widely used criterion that takes into account which of the phases (gas or liquid) is in a continuous or disperses form. Using this criterion, absorbers can be classified into the following groups:

3.6 Industrial absorbers

69

1. Absorbers in which both phases are in a continuous form (packed columns) 2. Absorbers with a disperse gas phase and a continuous liquid phase (plate columns, bubble columns, packed bubble columns) 3. Absorbers with a dispersed liquid phase and a continuous gas phase (spray columns) The industrially most frequently used absorbers are packed, plate, spray and bubble columns. 3.6.1 Packed columns Packed columns are the units most often used in absorption operations (Fig. 3.9). Usually, they are cylindrical columns up to several meters in diameter and over 10 meters high. A large gas-liquid interface is achieved by introducing various packings into the column. The packing is placed on a support whose free cross section for gas flow should be at least equal to the packing porosity. Liquid is fed in at the top of the column and distributed over the packing through which it flows downwards. To guarantee a uniform liquid distribution over the cross-section of the column, a liquid distributor is employed. Gas flows upward countercurrently

Fig. 3.9 Schematic and operating principles of packed, tray and spray towers (adapted from [18]).

70

3 Absorption and stripping

to the falling liquid, which absorbs soluble species from the gas. The gas, which is not absorbed, flows away from the top of the column, usually through a mist eliminator or demister. The mist eliminator separates liquid drops entrained by the gas from the packing. The separator may be a layer of the packing, mesh or it may be specially designed. Packings may be divided into two main groups: random and structured. The most popular random packing is rings. Raschig rings have a large specific packing surface and high porosity. They are hollow cylinders with an external diameter equal to the ring height that can be made of ceramic, metal, graphite or plastic. Recently, the application of structured packing in absorption has increased rapidly owing to the fact that it has a larger mass transfer area than random packing. Another advantage is that by using this packing it is possible, in contrast to random packing, to obtain the same values of mass transfer coefficient in the entire column. 3.6.2 Plate columns Another basic type of equipment widely applied in absorption processes is a plate column (Fig. 3.9). The diameter and height of the column can reach 10 and 50 m, respectively, but usually they are much smaller. In plate columns various plate constructions guarantee good contact of the gas with the liquid. Taking into account the whole column, the flow of the gas with the liquid has a countercurrent character. Liquid is supplied to the highest plate, flows along it horizontally and, after reaching a weir, flows through a downcomer from plate to plate in cascade fashion. Gas is supplied below the lowest plate, and then it flows through perforations in the plate dispersers (e.g., holes in a sieve tray, slits in a bubble-cap tray) and bubbles through the flowing liquid at each plate. The application of such a flow pattern is aimed at ensuring the maximum mass transfer area and high turbulence of the gas and liquid phases, which results in obtaining high mass transfer coefficients in both phases. The distances between plates should be such that liquid droplets entrained by the gas are separated from the liquid and the gas is separated from the liquid in the downcomer. Usually, the plate-to-plate distance ranges from 0.2 to 0.6 m and depends mainly on the column diameter and the liquid load of the plate. 3.6.3 Spray and bubble columns In spray towers liquid is sprayed as fine droplets, which make contact with a cocurrently or counter-currently flowing gas (Fig 3.9). The gas and liquid flows are similar to plug flow. Bubble columns (Fig. 3.10) are finding increasing application in processes when absorption is accompanied by a chemical reaction. They are also widely used as chemical reactors in processes where gas, liquid and solid phases are involved. In bubble columns the liquid is a continuous phase while the

3.6 Industrial absorbers

Fig. 3.10

71

Schematic of a bubble column absorber, without and with internal packing.

gas flows through it in the form of bubbles, the character of the gas flow is well represented by plug flow, while that of the liquid is between plug and ideally mixed flow. These absorbers can operate in cocurrent and countercurrent phase flow. 3.6.4 Comparison of absorption columns In each apparatus, due to different hydrodynamic conditions, various values of the mass transfer coefficients occur in both phases. Therefore, when choosing a given type of absorber, the following criteria should be taken into account: 1. the required method of absorption (continuous, semi-continuous) 2. the flow rate of the gas and the liquid entering the absorber (e.g. high gas flow rate and low liquid flow rate need different types of equipment than in the case where the two flow rates are commensurate) 3. the required liquid hold-up (large or small liquid hold-up is needed) 4. which phase controls mass transfer in the absorber (gas phase or liquid phase) 5. is it necessary to remove heat from the absorber? 6. the corrosiveness of the absorption systems 7. the required size of the interface (large or small mass transfer area) 8. the physicochemical properties of the gas and the liquid (particularly viscosity and surface tension).

72

3 Absorption and stripping

Apart from these factors there are many others, which influence the selection of equipment. For instance, gas impurities or deposits formed during the absorption process require the application of a given absorber. Packed columns are preferable to tray columns for small installations, corrosive service, liquids with a tendency to foam, very high liquid-to-gas ratios and low pressure drop applications. In the handling of corrosive gases, packing but not plates, can be made from ceramic or plastic materials. Packed columns are also advantageous in vacuum applications because the pressure drop, especially for regularly structured packings, is usually less than in plate columns. In addition packed columns offer greater flexibility because the packing can be changed with relative ease to modify column-operating characteristics. Tray columns are particularly well suited for large installations and low-tomedium liquid flow rate applications. In general they offer a wider operating window for gas and liquid flow than a countercurrent packed column. That is, they can handle high gas flow rates and low liquid flow rates that would cause flooding in a packed column. Tray columns are also preferred in applications having large heat effects since cooling coils are more easily installed in plate towers and liquid can be withdrawn more easily from plates than from packings for external cooling. Furthermore they are advantageous for separations that require tall columns with a large number of transfer units because they are not subject to channeling of vapor and liquid streams which can cause problems in tall packed columns. The main disadvantages of tray columns are their high capital cost, especially when bubble-cap trays or special proprietary design are used, and their sensitivity to foaming. Spray contactors are used almost entirely for applications where pressure drop is critical, such as flue gas scrubbing. They are also useful for slurries that might plug packings or trays. Other important applications include particulate removal and hot gas quenching. When used for absorption, spray devices are not applicable to difficult separations because they are limited to only a few equilibrium stages even with countercurrent spray column designs. The low efficiency of spray columns is believed to be due to entrainment of droplets in the gas and back mixing of the gas induced by the sprays. The high energy consumed for atomizing liquid and liquid entrainment in the gas outlet stream are two additional important disadvantages. Bubble columns are particularly well suited for applications when significant liquid hold-up and long liquid residence time are required. An advantage of these columns is their relatively low investment cost, a large mass transfer area and high mass transfer coefficients in both phases. Disadvantages of bubble columns include a high-pressure drop of the gas and significant back mixing of the liquid phase. The latter disadvantage can be reduced by introducing an inert packing of high porosity to the bubble column. Such a packing eliminates to a large extent the effects of liquid phase mixing along the column height. In addition the packing may also cause an increase in the mass transfer surface area with relation to the

Exercises

73

bubble column at the same flows of both phases. The advantages of bubble columns appreciably exceed their disadvantages and therefore an increasing number of these columns are applied in industry.

Nomenclature A f G H ΔHvap K KA L N p P0 Ptot R S T x, y X, Y α γ

L /K · G, absorption factor, Eq. 3.22 fraction absorbed or stripped (Eqs. 3.29 and 3.34) gas flow Henry coefficient molar heat of vaporization equilibrium constant, Eq. 3.21 distribution coefficient, Eq. 3.3 liquid flow number of (theoretical) stages partial pressure saturation pressure total pressure gas constant K · G / L, stripping factor, Eq. 3.30 temperature mole fraction mole ratio selectivity, relative volatility, equilibrium constant activity coefficient

[–] [–] [mol s–1] [–] [J mol–1] [–] [–] [mol s–1] [–] [N m–2] [N m–2] [N m–2] [J mol–1 K–1] [–] [K] [–] [–] [–] [–]

Indices A n ts vap

components stage number theoretical stages evaporation

Exercises 1. A plate tower providing six equilibrium stages is employed for stripping ammonia from a waste water stream by means of countercurrent air at atmospheric pressure and 25 °C. Calculate the concentration of ammonia in the exit water if the inlet liquid concentration is 0.1 mole% ammonia in water, the inlet air is free of ammonia and 2000 standard cubic meter (1 atm, 25 °C) of air are fed to the tower per m3 of waste water. The absorption equilibrium at 25 °C is given by the relation yNH3 = 1.414 xNH3 . Mwater = 0.018 kg mol–1; ρwater = 1000 kg m–3; R = 8.314 J mol–1K–1; 1 atm = 1.01325 bar. 2. When molasses is fermented to produce a liquor containing ethanol, a CO2-rich vapor containing a small amount of ethanol is evolved. The alcohol can be recovered by absorption with water in a sieve tray tower. For the following conditions, determine the number of equilibrium stages required for countercurrent flow of liquid and gas, assum-

74

3 Absorption and stripping

ing isothermal, isobaric conditions in the tower and neglecting mass transfer of all components except ethanol. Entering gas: 180 kmol/h, 98 % CO2, 2 % ethyl alcohol, 30 °C, 1.1 bar Entering absorbing liquid: 100 % water, 30 °C, 1.1 bar Required ethanol recovery: 95 % The vapor pressure of ethanol amounts 0.10 bar at 30 °C, and its liquid phase activity coefficient at infinite dilution in water can be taken as 7.5. 3. A gas stream consists of 90 mole% N2 and 10 mole% CO2. We wish to absorb the CO2 into water. The inlet water is pure and is at 5 °C. Because of cooling coils the operation can be assumed to be isothermal. Operation is at 10 bar. If the liquid flow rate is 1.5 times the minimum liquid flow rate, how many equilibrium stages are required to absorb 92 % of the CO2. Choose a basis of 1 mole/h of entering gas. The Henry coefficient of CO2 in water at 5 °C is 875 bar. 4. A vent gas stream in your chemical plant contains 15 wt% of a pollutant, the rest is air. The local authorities want to reduce the pollutant concentration to less than 1 wt%. You have decided to built an absorption tower using water as the absorbent. The inlet water is pure and at 30 °C. the operation is essentially isothermal. At 30 °C your laboratory has found that at low concentrations the equilibrium data can be approximated by y = 0.5 · x (where y and x are weight fractions of the pollutant in vapor and liquid). Assume that air is not soluble in water and that water is nonvolatile. a. Find the minimum ratio of water to air (L ′ / G ′)min on a solute-free basis b. With (L ′ / G ′) = 1.22 (L ′ / G ′)min find the total number of equilibrium stages and the outlet liquid concentration 5. A gas treatment plant often has both absorption and stripping columns as shown in Fig. 3.11. In this operation the solvent is continually recycled. The heat exchanger heats the saturated solvent, changing the equilibrium characteristics of the system so that the solvent can be stripped. A very common type of gas treatment plant is used for the drying of natural gas by physical absorption of water in a hygroscopic solvent such as diethylene glycol (DEG). In this case dry nitrogen is used as the stripping gas. a. At a temperature of 70 °C and a pressure of 40 bar the saturated vapor pressure of water is equal to 0,2 bar. It is known that water and DEG form a nearly ideal solution. Calculate the vapor-liquid equilibrium coefficient and draw the equilibrium line. b. Construct the operating line for xin = 0.02, yout = 0.0002 and L / G = 0.01. Determine the number of stages required to reduce the water molefraction from yin = 0.001 to yout = 0.0002. c. How many stages are required for L / G = 0.005. What happens for L / G = 0.004. Determine the minimal L / G ratio to obtain the desired separation. d. Desorption takes place at 120 °C and 1 bar. The saturated vapor pressure of water is equal to 2 bar. Construct the equilibrium and operating lines for desorption using yin = 0 and L / G = (L / G)max / 1.5. xout and xin are to be taken from the absorber operating at L / G = 0.01. e. Calculate analytically the number of stages in both sections. f. Comment on the chosen value of the liquid mole fraction at the outlet of the absorber.

Exercises

Fig. 3.11 Schematic of natural gas absorptive drying operation.

75

4 General design of gas/liquid contactors

4.1 Introduction In the previous chapters it was shown how to calculate the number of theoretical stages to separate a given feed with two components of different volatility into two fractions of predefined composition. Distillation is usually conducted in vertical cylindrical vessels that provide intimate contact between the rising vapor and the descending liquid. A distillation column normally contains internal devices for such an effective vapor-liquid contact. Their basic function is to provide efficient mass-transfer between the two-phase vapor-liquid system. However, the short contact times between the two phases in distillation does not allow complete establishing of thermodynamic equilibrium. The first aim of this chapter is to derive a tool to estimate the column height for a given separation. The number of real stages determines the height of the column. Therefore, we need to develop a model that translates the distance to the thermodynamic equilibrium to that number of real stages. The knowledge of the rate of interface mass-transfer will appear to be very helpful in this matter. The second aim of this chapter is to estimate the minimum diameter of the column to be designed for a certain capacity. Both goals are achieved by modeling the vapor-liquid contact, starting from a kinetic analysis of mass-transfer between the two phases involved. To understand what factors influence the rate of interface mass-transfer, we first derive rate expressions for such mass-transfer in idealized hydrodynamic environments. Although parts of the models derived below are applicable to liquid-liquid extraction, the examples will focus on two important devices for distillation, absorption and stripping: the plate column and the packed column.

4.2 Modeling mass-transfer In gas-liquid contactors two phases are present, separated by an interface in between. Assume a component A is diffusing from the gas phase to the liquid phase due to a difference in mole fraction as the driving force. Generally, diffusion as well as convection contributes to mass transport. Neglecting any other contribution, the rate of mass-transfer of a component A at some distance z from the interface is given by

78

4 General design of gas/liquid contactors V ΦA = − DAB Az ρV

dyA dz

+ yA (ΦA + ΦB) Az

(4.1a)

+ xA (ΦA + ΦB) Az

(4.1b)

and L Az ρL ΦA = − DAB

where ΦA ρV , ρL V L , DAB DAB Az yA , xA z

dxA dz

= molar transfer rate of A = molar density of vapor and liquid, resp. = coefficient of molecular diffusion in vapor and liquid, resp. = area at distance z from interface = mole fraction of A in vapor and liquid, resp. = distance to interface

[mol s–1] [mol m–3] [m2 s–1] [m2] [–] [m]

A mole fraction profile in case of spherical liquid or vapor droplets could be as shown in Fig. 4.1a. In the stationary situation, the gradient shown is inversely proportional to the distance z squared1. It is very reasonable to assume that there would be no resistance to transfer across the interface2, hence yAi = K · xAi . Then,

Fig. 4.1 Mole fraction profiles around spherical liquid or vapor droplets. (a) actual profiles; (b) profiles according to film theory.

1

See exercise 4.1. Except for very fast reactions at the interface, see R. B. Bird, W. E. Stewart & E. N. Lightfoot, Transport Phenomena, 2nd Edition, John Wiley & Sons, 2007.

2

4.2 Modeling mass-transfer

79

at low concentration (xA ≪ 1) or with equimolecular diffusion (ΦA = ΦB), the rate expressions given in Eq. 4.1 reduce to V ΦA = − DAB ALV ρV

dyA dz

|

L = − DAB ALV ρL

dxA dz

z=0

|

(4.2) z=0

where ALV = interfacial area between liquid and vapor, comprising the total surface area of all droplets. This area depends on the droplet size, dp . For n vapor droplets in a total volume V, the ratio of interfacial area ALV and a vapor volume3 εV · V reads (compare Eq. 6.1) ALV εV ⋅ V

=

n ⋅ π dp2 n⋅

π 3 dp 6

=

6 dp

A useful simplification is achieved by adopting the film mass-transfer theory to relate the driving force and the transfer rate. This theory is based on two assumptions (see Fig. 4.1b): – a thin, laminar film exists at either side of the interface, and – in this film only molecular diffusion is taking place. Hence, according to this theory, no gradients exist in the bulk of the two phases. Resistance to mass-transfer is concentrated in the boundary layers where only molecular diffusion is occurring. The stationary state rate expressions according to the film theory are: ΦA = kV ALV ρV (yA − yAi) = kL ALV ρL (xAi − xA)

(4.3)

where kV , kL = film or single-phase mass-transfer coefficients [m s–1]. When the boundary layers are very thin, the gradients in Eq. 4.2 can be replaced by their difference quotients. Now, comparison with Eq. 4.3 reveals that the masstransfer coefficient, k, equals the ratio of molecular diffusivity and film thickness, DAB /δ, where δ = Δz. Literature data of molecular diffusivities in the gas phase show values in the order of 10–5 – 10–4 m2 s–1, whereas mass-transfer coefficients appear to be in the order of 0.01 m s–1. Thus, the order of magnitude of the film thickness in the gas phase, δV , is 10–3 – 10–2 m. Diffusivities in the liquid phase are much smaller, in the order of 10–9 m2 s–1, experimental values of liquid mass-transfer coefficients are around 10–4 m s–1. Thus, the film thickness in the liquid phase, δL , is orders of magnitude smaller than that in the gas phase. In either case, the film thickness is sufficiently small to assume linear concentration gradients. The simplified picture according to the film theory is shown in Fig. 4.1b. This gives only a rough, inaccurate model of transport through liquid vapor 3

εV = vapor volume fraction.

80

4 General design of gas/liquid contactors

Fig. 4.2 Driving force and interface compositions. (a) Both transport in liquid and vapor determine overall transport rate; (b) transport in vapor is very fast; (c) transport in liquid is very fast.

interfaces. Nevertheless, the film theory appears to be an adequate tool for many engineering applications, especially when it comes to understanding principles of mass-transfer in absorption or distillation towers. The interface compositions xAi and yAi , shown in Fig. 4.1, are connected through Eq. 4.3, which can be rearranged to show as a linear operating line in a y-x diagram: yA − yAi = −

kL ρL kV ρV

(xA − xAi)

(4.4)

The slope of this straight line depends on the ratio of the two mass-transfer coefficients involved. The absence of mass-transport resistance in the interface suggests that both interface compositions are in equilibrium. This assumption gives the second equation required to calculate the interface compositions. A graphical method to do so is presented in Fig. 4.2. Eq. 4.4 and the equilibrium curve are plotted in a y-x graph. The intersection of the straight line with the equilibrium curve gives the values of xAi and yAi . Two limiting situations can be considered. When the transport in the gas phase is very fast compared to that in the liquid phase, the gradient across the gas film can be neglected to that across the liquid film. All resistance to mass transport is said to be in the liquid phase; this situation is depicted in Fig. 4.2b. Fig. 4.2c shows the opposite situation where the resistance is entirely in the gas phase. The application of the above method to model transfer rates in G/L contactors would be much easier to handle when the interface compositions could be eliminated from the driving force factor. The following mathematical treatment will do just that4. For the sake of simplicity we assume a linear equilibrium expression: yA* = K ⋅ xA 4

See also exercise 4.2.

(2.8)

4.2 Modeling mass-transfer

81

where yA* is a hypothetical gas phase composition in equilibrium with the actual mole fraction xA in the liquid phase (see first diagram on page 81). Combination with Eq. 4.3 gives yA − yAi =

ΦA

and

kV ALV ρV

K ⋅ xAi − K ⋅ xA = yAi − yA* =

K ⋅ ΦA kL ALV ρL

(4.5)

Summation of both expressions eliminates yAi giving: ΦA = kOV ALV ρV (yA − yA*)

(4.6)

with kV

kOV = 1+

kV KρV ⋅ kL ρL

=

kV 1+

mkV kL

kOV = overall mass-transfer coefficient, based on the gas phase [m s–1] m=

KρV ρL

, volumetric distribution coefficient [–]

(see Eq. 4.6)

(see Eq. 4.7)

82

4 General design of gas/liquid contactors

Similarly, elimination of xAI from Eq. 4.3 gives a rate expression with an overall mass-transfer coefficient kOL , applicable when the main resistance against mass transfer is in the liquid phase (see second diagram on page 81): ΦA = kOL ALV ρL (xA* − xA)

(4.7)

with kOL =

kL 1+

kL m kV

Summarizing, the rate of mass-transfer across a phase boundary depends on – the driving force, yA − yA* or xA* − xA, the asterisk * referring to a hypothetical equilibrium – the contact area ALV between the two phases, and – an overall mass-transfer coefficient, kOV or kOL . The maximum mass-transfer rate is obtained when these three factors on the righthand side of the Eqs. 4.6 or 4.7 are as large as possible: • The largest driving force is achieved when the overall flow pattern allows countercurrent contact between equal streams of gas and liquid without significant remixing. An important condition is that both phases are distributed uniformly over the entire flow area. • A large contact area ALV is desirable for mass-transfer and mainly determined by the used column internals. Depending upon the type of internal devices used, the contacting may occur in discrete steps, called plates or trays, or in a continuous differential manner on the surface of a packing material. Tray columns and packed columns are most often used for distillation since they guarantee excellent countercurrent flow and permit a large overall height. The internals provide a large mass-transfer area. In tray columns mass-transfer areas range from 30 to 100 m2 per unit tray area5, in packed columns from 200 to 500 m2 per unit column volume. • The mass-transfer coefficient increases proportionally with the relative velocities between the liquid and vapor phases and is also improved by constant regeneration of the contact area between the phases. It is difficult to measure overall mass-transfer coefficients as such. The usual experimental set-ups designed to determine mass transport under well-defined conditions (temperature, pressure, concentration, hydrodynamic flow pattern) produce data to calculate the product of mass-transfer coefficient and contact area. 5

Specific interfacial area ALV in gas-liquid systems may be estimated from gas volume fraction ε and bubble diameter d as ALV = 6 ε/d in m2/m3 (see Eq. 4.2, see also Eq. 6.2 for interfacial area in cylinders).

4.3 Plate columns

83

In the following sections L/G mass transfer will be modeled in tray – or plate − columns as well as in packed columns.

4.3 Plate columns Fig. 4.3 shows the most important features of a plate or tray column. The gas flows upwards within the column through perforations in horizontal trays, and the condensed liquid flows countercurrently downwards. However, as indicated in Fig. 4.4,

Fig. 4.3 Cutaway section of a plate column (a) downcomer; (b) tray support; (c) sieve trays; (d) man way; (e) outlet weir; (f) inlet weir; (g) side wall of downcomer; (h) liquid seal. (Reproduced with permission from [26].)

Fig. 4.4 Schematic of flow pattern in a cross-flow plate distillation column.

84

4 General design of gas/liquid contactors

the two phases exhibit cross flow to each other on the individual trays. The liquid enters the crossflow plate from the bottom of the downcomer belonging to the plate above and flows across the perforated active or bubbling area. The ascending gas from the plate below passes through the perforations and aerates the liquid to form a large interfacial area between the two phases. It is in this zone where the main vapor-liquid mass transfer occurs. The vapor subsequently disengages from the aerated mass on the plate and rises to the tray above. The aerated liquid flows over the exit weir into a downcomer where most of the trapped vapor escapes from the liquid and flows back to the interplate vapor space. Some of the liquid accumulates in each downcomer to compensate for the pressure drop caused by the gas as it passes through the tray. The liquid then leaves the plate by flowing through the downcomer outlet onto the tray below. In large diameter crossflow plates, multiple liquid-flow-path plates with multiple downcomers are used to prevent the hydraulic gradient of the liquid flowing across the plate becomes excessive. Three principal vapor-liquid contacting devices are used for cross-flow plate design: the bubble cap tray, the sieve tray and the valve tray, see Fig. 4.5: • Bubble cap trays have been used almost exclusively in the chemical industry until the early 1950s. As shown in Fig. 4.5, their design prevents liquid from leaking downward through the tray. The vapor flows through a hole in the tray floor, through the riser, reverses direction in the dome of the cap, flows downward and exits through the slots in the cap. However, the complex bubble caps

Fig. 4.5 Schematic of a bubble cap, sieve and valve tray.

4.3 Plate columns

85

Fig. 4.6 Examples of valves uses in valve trays.

are relatively expensive and have a higher pressure drop than the other two designs. This limits their usage in newer installations to low liquid flow rate applications or to those cases where the widest possible operating range is desired. • Sieve trays have become very important because they are simple, inexpensive, have high separation efficiency, and produce a low pressure drop across the tray. Conventional sieve trays contain typically 1–12 mm holes and exhibit ratios of open area to active area ranging from 1 : 20 to 1 : 7. If the open area is too small the pressure drop across the tray is excessive, if the open area is too large the liquid weeps or dumps through the holes. • Valve trays are a relatively new development that represents a variation of the sieve tray with liftable valve units such as those shown in Fig. 4.6 fitted in the holes. The liftable valves prevent the liquid from leaking at low gas loads and to avoid excessive pressure increase at high gas loads. The main advantage is the ability to maintain efficient operation while being able to vary the gas load up to a factor of 4 to 5. This capability gives valve trays a much larger operational flexibility than any other tray design.

4.3.1 Dimensioning a tray column Product specification of a distillation, absorption or stripping process includes at least product purity and capacity. In the previous chapters methods are discussed to determine the number of theoretical stages required to achieve desired product purity. This relates to the height of the column, the more trays that are required, the higher the column. The capacity of the process determines the diameter of the column. This section focuses on methods to estimate the minimum dimensions of a tray column: the minimum column height necessary to meet the desired product purity and the minimum diameter necessary to accommodate the required capacity (at constant pressure, see Fig. 2.3).

86

4 General design of gas/liquid contactors

4.3.2 Height of a tray column Although the column requirements are calculated in terms of theoretical or equilibrium number of stages, Nts , the design must specify the actual number of stages, Ns . The minimum height of a column is then calculated from the minimum distance Hspacing between adjacent trays and the number of actual stages: Hcolumn = Hspacing ⋅ Ns

(4.8)

Thus, an estimate of column height involves determination of the number of real stages, Ns , and the tray spacing, Hspacing . 4.3.2.1 Tray spacing The distance between two trays should be as small as possible but sufficiently large to prevent liquid droplets to reach the tray above. In industrial tray columns spacing from 0.15 m to 1.0 m is used. This value depends on vapor load and tray diameter. In columns with a diameter of 1 m or larger, a typical tray spacing of 0.3 to 0.6 m is found. A value of 0.5 m appears to be a reasonable initial estimate. In a detailed tray design this value is subject to revision, however, that part of the design is outside the scope of this textbook. 4.3.2.1 Tray efficiency Due to relatively high flow rates in gas-liquid towers, contact times on trays are not sufficient to establish thermodynamic equilibrium. A measure of the performance of a real tray in approaching equilibrium is the efficiency of a real tray. In the previous chapters methods were discussed to determine the total number of theoretical stages for a given separation such as distillation, absorption or stripping. The overall efficiency EO of a column is defined by the number of theoretical trays, Nts , divided by the actual number of trays required to obtain the desired product purity6 EO ≡

Nts Ns

(4.9)

Most hydrocarbon distillation systems in commercial columns achieve overall tray efficiencies of 60–80 %7, in absorption processes the range is 10–50 %. These figures are just guidelines, they have no theoretical basis and are not suitable for design purposes. A more fundamental approach would be to derive an expression for the efficiency of a single plate in terms of local driving force for interphase mass transport and hydrodynamic conditions. Firstly, we need a hydrodynamic 6

In distillation, where the reboiler counts as one theoretical stage, this definition should be adjusted accordingly. 7 EO sometimes is expressed as a percentage by multiplication with 100.

4.3 Plate columns

87

Fig. 4.7 Mass transport on a model non-equilibrium tray in absorber. (a) distance to equilibrium; (b) elements in mass balance for interphase mass transport.

model of a tray. A reasonable model would be to assume that the gas phase, passing at high velocity through the liquid layer on a tray, behaves as a plug flow. Further that its concentration changes only in axial direction, radial concentration gradients can be neglected. The gas flow causes extensive mixing in the liquid layer, so the liquid is assumed being ideally mixed. The composition yn in gas phase leaving the nth tray is independent of the radial position. Now the tray, plate or Murphree efficiency EMV compares the actual difference yn + 1 − yn leaving two consecutive trays to the maximum achievable difference yn + 1 − yn* : EMV =

yn + 1 − yn

(4.10)

yn + 1 − yn* Fig. 4.7 gives an amplification of the tray efficiency in an absorber. Here, the lighter component in the gas phase from tray n + 1 is absorbed by the liquid, which has the same composition xn everywhere on the tray. At increasing height in the liquid layer, the mole fraction y decreases. On an equilibrium tray the composition of the gas phase leaving tray n would equal yn* whereas on a real tray the exit value amounts to yn . In the following paragraph the concept of transfer units is used to estimate the number of real trays, Ns . 4.3.2.3 Transfer units The next step is to relate the mole fractions in Eq. 4.10 to kinetic parameters and flow rates. The hydrodynamic simplification for a plate defined above (liquid ideally mixed, plug flow in gas phase) allows setting up a mass balance in a thin liquid layer at a certain position h above the tray. Instead of the height h, it is

88

4 General design of gas/liquid contactors

convenient to take the interphase ALV present on the tray as independent parameter. A thin liquid layer with thickness Δh comprises ΔA unit surface area. The driving force for interphase transport at this position equals y − yn*. Note that yn*, the virtual gas phase composition in equilibrium with the liquid of composition xn , is constant. In the stationary situation the difference in molar flow of the absorbable component in the gas phase is balanced by interphase transport, see Fig. 4.7. Applying the rate expression given in Eq. 4.6 the mass balance becomes: V ⋅ y = V ⋅ (y + Δy) + kOV ⋅ ρV ⋅ (y − yn*) ⋅ ΔA

(4.11)

For the sake of clarity, we restrict ourselves to the case that heat effects can be neglected and that the molar vapor flow rate, V, is constant, e.g. equal molal overflow and/or the concentration is low (less than 10 mole%). Taking the differential and integrating from y = yn+1 at the liquid – tray interphase where interface surface area A = 0 to y = yn at the top of the liquid layer where total LV-surface area A = ALV, it follows that yn

∫

dy

* yn + 1 yn

−y

≡ NOV = ln

yn + 1 − yn* yn − yn*

=

kOV ⋅ ALV V ρV

(4.12)

The left-hand side of Eq. 4.12 represents the ratio of actual change in composition to driving force and characterizes the difficulty of separation. This dimensionless quantity is called the overall number of gas-phase transfer units NOV. The larger the number of transfer units, the more efficient the tray. Combination of Eqs. 4.10 and 4.12 relates the number of transfer units to the Murphree efficiency, EMV : EMV = 1 − e −NOV

(4.13)

This expression for the Murphree efficiency holds for mass transfer between a vapor phase plug flow and an ideally mixed liquid phase and is applicable to all three types of trays. The right-hand side of Eq. 4.12 contains factors that depend on tray dimensions and operating conditions, such as liquid and vapor load. It can be interpreted as the ratio of interphase mass transport k · A [m3s–1] to convective vapor flow Q = V/ρV [m3s–1]. This ratio is connected to the height of a transfer unit, Htu NOV =

kOV ⋅ ALV QV

=

kOV ⋅ Ah QV

⋅H ≡

H Htu

(4.14)

where Ah = interfacial area per unit length and H = height of liquid layer. In a plate column, Htu has little meaning for the determination of the real number of stages. In section 4.4 the concept of the height of a transfer unit will be applied to estimate the height of a packed column.

4.3 Plate columns

89

4.3.2.4 Overall efficiency Finally, the overall efficiency EO in the column should be related to the tray efficiency EMV . We have to find the number of real trays with efficiency EMV that achieve the same change in composition as with Nth theoretical trays. A graphical method would be to replace the equilibrium line for theoretical stages in a McCabe-Thiele plot with a pseudo-equilibrium line for real stages as shown in Fig. 2.20. Real stages are stepped off between this pseudo-equilibrium line and the operating line. Keep in mind that the reboiler in a distillation column is considered to count as one theoretical stage. If the operating and equilibrium lines are straight, an analytical expression is more convenient. In the y-x plot of Fig. 4.8 step ABC represents an equilibrium tray with an arbitrary equilibrium change in vapor composition Δy1. ADE and EFG illustrate the effect of two non-equilibrium trays on vapor composition, Δy1′ and Δy2′, respectively. By definition, AD/AB, EF/EJ etc. are the tray efficiencies EMV . At constant values of EMV , D and F are points of a straight pseudo-equilibrium line. In this case, the ratio of change in composition of two consecutive trays is constant: Δy2′ Δy1′

=

EF AD

=

EF EJ (JH + HE) ⋅ = EMV ⋅ EJ AD AD

(4.15)

but JH = K ⋅ HB =

KV L

⋅ AD

and HE = AB − AD =

1

(EMV

− 1 ⋅ AD,

)

thus the ratio

Δy 2′ Δy 1′

is constant: Δy 2′ Δy 1′

= 1 + EMV (S − 1) = s

(4.16)

with separation factor S = SS = KV/L for stripping (as in Fig. 4.8) and S = SA= L /KV for absorption. The total change in composition achieved by N trays, each with efficiency EMV ≡ Δy1′ /Δy1 equals: N ′ 2 N−1 ∑ Δyi = Δy1 ⋅ EMV ⋅ (1 + s + s + ... + s ) = Δy1 ⋅ EMV ⋅

sN − 1

(4.17) s−1 This result would be the first step to derive a general, non-equilibrium Kremser equation that transforms into the equilibrium Kremser equations 3.29 or 3.34 for 1

90

4 General design of gas/liquid contactors

Fig. 4.8 Construction of real, non-equilibrium trays subject to mass transport resistance. ABC represents an equilibrium plate; ADE and EFG represent real trays with constant tray efficiency EMV .

EMV → 1. Here, we just need Eq. 4.17, which relates the number of trays and the total composition change in one phase. Ns real trays should give the same total change in composition as Nth theoretical plates with EMV = 1 and s = S. Application of Eq. 4.17 to both types of trays gives: Δy1 ⋅ EMV ⋅

s Ns − 1 s−1

= Δy1 ⋅

S Nth − 1 S−1

(4.18)

and, after rearranging EO =

Nth Ns

=

ln [1 + EMV (S − 1)] ln (S)

(4.19)

Now the overall efficiency EO and the real number of stages Ns can be calculated from the separation factor and the tray efficiency. With the initial estimate of tray spacing and the number of real stages, the column height can be calculated from Eq. 4.8. It should be noted that the hydrodynamic model used above is an oversimplification. Especially on larger trays, the assumption of complete mixing is far from realistic. In fact, cross-flow is more likely, as shown in Fig. 4.4. Liquid flows from the downcomer clearance across the plate and leaves it via the exit weir. The liquid entering the plate is richer in the volatile component than is the liquid leaving the plate. Vapor enrichment is higher than with complete mixing,

4.3 Plate columns

91

where the liquid composition anywhere on the tray equals the lower exit composition. Consequently, tray efficiency in cross-flow is higher than with complete mixing. Cross-flow models would be more appropriate to quantify this effect and are available in literature (King [5]). However, for a basic understanding of the phenomena determining the column size, the simple mixing-plug flow model is adequate. 4.3.3 Diameter of a tray column In the previous section methods are discussed to estimate column height for a given separation from first principles in thermodynamics and kinetics. Any approach involves calculation of the slope of an operating line, L/V. The molar flow rates of liquid, L and vapor, V, are also of great importance in estimating a column diameter. More detailed column dimensioning requires determination of the entire operating region bound by a range of liquid and vapor flow rates as shown in Fig. 4.9. Here we limit ourselves to a discussion of the influence of flow rates on performance, only to illustrate the most important phenomena occurring on a tray. More attention will be paid to the maximum vapor load of the operating region because, at a given vapor volume flow rate QV in [m3s–1], the minimum column diameter is determined by the maximum allowable vapor flow rate uV,max in [ms–1]. Fig. 4.9 shows several mechanisms that limit the operation range of a sieve plate. Boundaries to the region of stable operation exist at minimum and maxi-

Fig. 4.9 Schematic operating diagram for cross-flow plates.

92

4 General design of gas/liquid contactors

mum vapor and liquid loads. The weeping (or dumping) line represents the minimum operable vapor flow rate at various liquid flow rates. Below the line, the vapor rate is too low to maintain the liquid on the plate and the liquid weeps, or even dumps, through the plate orifices. At low liquid loads liquid droplets may be carried with the vapor to the tray above. This entrainment lowers the tray efficiency because liquid of lower volatility is recycled to a tray of higher volatility. Coning occurs, at higher vapor loads, when the vapor pushes aside the liquid from the orifices. This prevents a good interphase contact, also decreasing the plate efficiency. To minimize this effect, exit weirs (see Fig. 4.3) should be high enough to ensure sufficient liquid on the tray. Flowing of the vapor upward through the downcomers, bypassing the tray, is called blowing, which may also occur at low liquid loads, especially if the downcomer slit is large. Extremely low liquid loads cause an uneven liquid distribution across the tray also decreasing mass transfer efficiency. These lower limits are not absolute, although below these limits separation efficiency becomes worse. Unlike these lower limits, the upper limits discussed below are absolute. Above these upper limits, column operation stops. At a high liquid load, especially with small downcomer areas, choking may occur: the liquid in the downcomer drags vapor down to the lower tray. Downcomer flooding, also called overflowing, occurs when the liquid flow through the column becomes larger than the downcomer capacity. The liquid fills the downcomer and recycles back to the previous tray with the rising vapor and proper countercurrent column operation breaks down. High vapor rates increase pressure drop across the tray, encouraging flooding. At high vapor loads, liquid can be blown off the tray in the form of fine droplets. This entrainment flooding prevents the liquid from flowing counter currently to the vapor. Again, proper column operation stops. 4.3.3.1 Flooding Apart from geometrical ratios considering tray configuration, six factors can be identified as having a major importance considering these limits. These six are: superficial vapor and liquid velocity, uV and uL, vapor and liquid density8 ρV and ρL, tray spacing Hspacing and gravitational constant g. All six factors determine the pressure drop across a hole in a tray. From dimensional analysis it appears that these six factors can be reduced to two dimensionless numbers: a gas number ρV ρL uL uV and a flow parameter . In literature, flooding properties ρL uV √ρV √ gH √ are given in charts where gas numbers are usually plotted as a function of a flow QL ρL parameter taking the form . QV and QL are volume flow rates [m3s–1] QV √ρV 8

Note that the dimensions of ρL and ρV are in kg m–3.

4.3 Plate columns

93

obtained from the molar vapor and liquid flows emerging from the stage-wise calculations discussed in the previous two chapters. The flow parameter represents the ratio of kinetic energies of the two phases and indicates the type of the two-phase mixture on a tray. A high value suggests a bubbly liquid with a dispersed gas phase, which is characteristic for absorption columns at high pressure. In such a system, the vapor rate can be increased until the point where the pressure drop across the tray balances the pressure drop in the downcomer. A higher liquid load starts downcomer flooding. A low value of the flow parameter indicates a system where the vapor phase is continuous, with dispersed liquid, as found in vacuum distillation. In this region, too high a value of vapor rate causes entrainment flooding and liquid droplets are blown to the tray above. In the following a simple model for each of the two types of flooding, downcomer and entrainment flooding resp., will be derived, predicting the maximum allowable vapor rate to avoid flooding. This maximum flow rate determines the minimum tray diameter. These derivations use two geometric ratios: ϕ=

Aholes Abubble

and

σ=

Aslit

(4.20)

Abubble

ϕ is the fraction hole area relative to the active or bubbling area Abubble of the tray. Abubble is that part of the tray not blocked by the downcomers. Aslit is the vertical surface area of the slit at the exit of the downcomer, equal to the product of the width of the slit and the downcomer clearance (the height of the slit, see Fig. 4.4). 4.3.3.2 Downcomer flooding At high liquid loads, the vapor rate can be increased to a point where the pressure drop across the orifices or holes equals that across a completely filled downcomer. Any further increase would overflow the downcomer, backing up the liquid to the tray above. A fully loaded downcomer contains liquid as well as a turbulent two-phase mixture on top. In the pressure balance below it is assumed that the effective liquid head is just half of the tray spacing. Reminding that uL uV and neglecting the contribution of the liquid layer on the uL, slit = , uV, hole = σ ϕ tray to the pressure drop, a stationary state pressure balance gives: Δp = Δp1 + Δp2 c1 ρV

uV, max 2

( ϕ )

or

= ρL g

Hspacing 2

− c2 ρL

uL 2

(σ)

(4.21)

The vapor pressure drop across the holes on the left hand side of Eq. 4.21 balances the downcomer pressure drop (static pressure by the liquid minus pressure drop across the slit) on the right-hand side. Rearrangement of Eq. 4.21 and replacing

94

4 General design of gas/liquid contactors

the velocity ratio by the volume flow ratio leads to the following downcomer flooding line equation: uV, flood √g ⋅ Hspacing

ρV

√ρL

=

ϕ √c1

0.5 1+

√

c2 ϕ 2 ρL Q L2

(4.22)

c1 σ 2 ρV Q V2

At these higher liquid loads, smaller fractional hole areas are preferable. These values are plotted in Fig. 4.10, for reasons of simplicity as a straight line ranging from ϕ = 6 % to ϕ = 10 %. In this plot, the two flooding lines, qualitatively depicted in Fig. 4.9, are now substantiated. The friction coefficients in Eq. 4.22 (c1 for holes and c2 for slits) have to be determined by experiments. Estimates can be obtained from first principles. In the turbulent conditions on a sieve tray, reasonable values for c1 and c2, are 1.1 and 1.6, respectively. To obtain the maximum allowable vapor rate, one more unknown in Eq. 4.22 has to be assigned a value to: σ, the ratio of vertical downcomer slit area and active tray area (Eq. 4.20). In practice a value between 0.01 and 0.05 is used. The gas number at flooding velocity is plotted in Fig. 4.10 as a function of the flow parameter. In a more detailed design the first estimate of σ

Fig. 4.10

Flooding properties of a sieve tray.

4.3 Plate columns

95

need to be optimized, which requires recalculating the maximum vapor rate from the adjusted downcomer flooding line, that will not discussed here. 4.3.3.3 Entrainment flooding The second mechanism occurring on a sieve tray, causing flooding by entrainment, involves tearing apart the liquid by the high kinetic energy of the vapor. The line describing entrainment flooding is not available as a function explicit in the flow parameter. Moreover, it contains all geometrical ratios of a sieve tray and therefore would require a detailed design of the tray, which is outside the scope of this textbook. In Fig. 4.10 we just give an entrainment flooding line for a non-specified tray without deriving the implicit function describing it. In the region where either mechanism might dominate, the lower gas number prevails. The thick parts of both flooding lines indicate this. The limiting value of the entrainment flooding line can be derived easily by noting that the Weber number for droplet formation relates the disruptive aerodynamic forces and the restoring surface tension forces: ρL d We =

uV 2 (ϕ ) (4.23)

γLV

where d = diameter of liquid droplet and γLV = surface tension [Nm–1]. At a vapor rate, just sufficient to start droplet formation, the critical value of the Weber number is approximately 12; at lower values surface tension restores the liquid surface. Once formed, the stationary falling rate of a droplet should at least equal the upward vapor velocity to avoid entrainment. In this limiting condition the stationary state force balance states that apparent weight equals resistance force: 1 6

3

π d ⋅ ( ρL − ρV) ⋅ g = C w ⋅

1 4

2

πd ⋅

1 2

ρV

uV, max

2

( ϕ )

(4.24)

For a sphere in the turbulent region, the friction factor Cw = 0.4. Elimination of the droplet diameter d from the latter two equations, applying the critical value of the Weber number at uV = uV, flood , leads to the following expression for the highest allowable gas number: uV, flood

ρV

√ρL √g Hspacing

= 2.51

[

ϕ 2 γLV (ρL − ρV) 2 ρL2 g Hspacing

0.25

]

(4.25)

The fractional hole area ϕ at the lowest liquid loading is read from the upper line in Fig. 4.10. The gas number thus calculated is shown as a horizontal square dotted line in Fig. 4.20.

96

4 General design of gas/liquid contactors

4.3.3.4 Tray diameter and pressure drop At this stage, the maximum allowable vapor rate to avoid flooding can be calculated at any given value of the flow parameter. The flooding lines in Fig. 4.10 are based on either Eq. 4.22 (downcomer flooding) or on Eq. 4.25 (entrainment flooding). The lowest value of the two is used for the diameter calculation. For safety reasons, a value of 85 % of the flooding velocity should be used while the downcomer area will occupy a further 15 % of the active tray (bubble) area. The active hole area and the minimum column diameter then follow from Abubble, min =

QV 0.85 ⋅ uV, flood

and

Dmin =

4 Abubble, min ⋅ 0.85 √π

(4.26)

Finally, the pressure drop across a sieve tray is calculated. The total pressure drop comprises two contributions, one due to the resistance in the holes (see Eq. 4.21) and another due to the static pressure of the liquid layer with height hliq on the tray: Δptray = Δpdry + Δpliq = c1 ⋅ ρV ⋅

0.85 ⋅ uV, flood

(

ϕ

2

)

+ ρL ⋅ g ⋅ hliq (4.27)

4.3.3.5 Remark on flooding charts Many authors use the flooding chart of Fair (as given in, e.g., Perryʼs Chemical Engineersʼ Handbook [16]), who plotted a non-dimensionless capacity factor Csb as a function of the flow parameter for various values of tray spacing. Vapor rate and capacity factor are related through: uV, flood = Csb, flood

γLV

0.2

ρL − ρV 0.5

(0.020) ( ρV )

(4.28)

In yet other published flooding charts the superficial vapor rate is plotted as a ρV uV function of superficial liquid rate, or as a gas number as a function √ρL √g H uL of liquid number . All these types of flooding diagrams are usually subject √g H to some restrictions: – – – – –

the vapor-liquid system is non-foaming; the weir height 15 % of plate spacing or less; sieve plate perforations are 13 mm or less in diameter; if not specified in the chart, the fraction hole area is 0.1; if not specified in the chart, the surface tension is 0.020 Nm–1.

4.4 Packed columns

97

4.4 Packed columns The most important features of a packed column are shown in Fig. 4.11. Obviously, the most important element of packed column internals is the packing itself. To promote mass transfer the packing should have a large surface area per unit of volume and be wetted by the liquid as completely as possible. The internals of the column must offer minimum resistance to gas flow. Modern packings have a relative free-cross-sectional area of more than 90 %. The vapor enters the bottom of the column and flows upward through the free cross-sectional area of the internals in countercurrent contact with the down flowing liquid. Because this

Fig. 4.11 Cutaway section of a packed column with a structured packing. (a) liquid distributor; (b) liquid collector; (c) structured packing; (d) support grid; (e) manway; (f) liquid redistributor. (Reproduced with permission from [26].)

98

4 General design of gas/liquid contactors

countercurrent flow exists throughout the column, packed columns are in principle more effective for mass transfer than a tray column. However, the countercurrent flow of gas and liquid in a packed column is not perfect since the liquid flow is not uniform over the cross section. Well-known mechanisms causing liquid maldistribution are liquid channeling and wall flow. Good contacting efficiency between the two phases is only obtained with a uniform liquid distribution over the entire cross-sectional area. For this purpose liquid distributors and redistributors are used. Redistributors are necessary to avoid the built up of a high degree of liquid maldistribution when the bed exceeds a height of 6 m. Other important supplementary elements, schematically depicted in Fig. 4.11, are necessary for proper column operation: • Liquid collectors are installed for the withdrawal of side stream products, pump arounds and the collection of liquid before each liquid distributor. • Wall wipers ensure that liquid accumulated on the wall is fed back into the packing. • Support grid is installed to support the packing and the liquid holdup of the packing. • Hold down plate has the primary function to prevent expansion of the packed bed as well as to maintain a horizontal bed level. • Gas distributor is used to obtain a uniform gas flow across the column cross section.

4.4.1 Random packing The type of packing used has a great influence on the efficiency of the column. Most industrial packed columns use random packings composed of a large number of specially formed particles. Presently, some 50 different types of random packings are offered on the market. Some examples of the more important types are shown in Fig. 4.12. The oldest packing element is the Raschig ring, with its

Fig. 4.12 Common random column packings.

4.4 Packed columns

99

characteristic feature that the length of the ring is equal to its diameter. This feature makes the particles form quite a homogeneous bed structure during pouring into the column. Raschig rings have a rather high pressure drop since the walls of those rings in horizontal position block the gas flow. The Pall ring avoids this disadvantage since parts of the wall are punched out and deflected into the inner part of the ring. A Pall ring has the same porosity and the same volumetric area as a Rashig ring but a considerably lower pressure drop. The feasible particle size depends on column diameter but should not exceed 1/10 to 1/30 of the column diameter. Most particles are made of metal or ceramics, but plastics are increasingly used. 4.4.2 Structured packing A certain degree of inhomogeneity is unavoidable in any random packing. These inhomogeneities, that cause liquid maldistribution, are avoided by using ordered packing structures such as the corrugated sheet structure that was developed by Sulzer in the mid sixties. As Fig. 4.13 shows the corrugated sheets are assembled parallel in vertical direction with alternating inclinations of the corrugations of neighboring sheets. Since the packing does not fit perfectly into the cylindrical column shell, additional tightening strips have to be installed between packing and column wall. These structured packings provide a homogeneous bed structure, and a low pressure drop due to the vertical orientation of the sheets. It is therefore not surprising that their first applications were found in vacuum distillation. At present structured packings are available in different types (gauze, sheet) and in a variety of materials (metals, plastics, ceramics, carbon). For most applications structured sheet metal packings offer a more attractive performance to cost ratio than structured gauze packings because the cost of structured sheet is about one-third that of gauze while the efficiencies are about the same. Disadvantages

Fig. 4.13

Schematic of corrugated structured packing (adapted from [26]).

100

4 General design of gas/liquid contactors

Fig. 4.14 Comparison of relative capacity and relative separation efficiency of various random and structured packings (adapted from [26]).

of structured packings are high costs relative to trays, criticality of initial liquid and vapor distributions and the associated hardware required. Typically the installed cost of structured sheet metal packing plus associated hardware is about three to four times that for conventional trays. A good packing design should have high capacity, high separation efficiency and large flexibility to gas and liquid throughput. Fig. 4.14 shows a rough comparison of capacity and separation efficiency of several metal packings. Structured packings are generally superior to random packings in both capacity and separation efficiency. Modern random packings are designed for low pressure drop and perform significantly better than the standard Rashig and Pall ring packings. 4.4.3 Dimensioning a packed column Like the design of tray columns, the design of packed columns starts with estimates of column height9, column diameter and pressure drop. The design method follows the same approach as that for tray columns. The optimal molar liquid and vapor flows, L and V, respectively, and the number of theoretical stages, Nts , are calculated at constant pressure following one of the methods discussed in the previous chapters. The height depends on the required product purity and is connected to the difficulty of separation. The diameter depends on the given capacity and the pressure drop follows from capacity, diameter and properties of the chosen packing.

9

With column height is actually meant the height of the packed section of the column.

4.4 Packed columns

101

4.4.4 Height of a packed column 4.4.4.1 Theoretical plates Basically, two different, but related approaches can be followed to obtain a first estimate of column height. In the first approach, the number of equilibrium stages determines the total height, Hcolumn : Hcolumn = Nts ⋅ Hets

(4.29)

using Hets which is the height equivalent to a theoretical stage; vapor and liquid leaving a section of height Hets of the column are in equilibrium. Hets is also known as HETP, the height equivalent of a theoretical plate. Various forms of empirical relations exist to find the HETP of packed columns. The following equation gives the HETP for packed columns using 25 and 50 mm Raschig rings: HETP = 18 dring + 12 K

V

(4.30)

(L − 1)

where dring = diameter of Raschig rings K = slope of the equilibrium line V, L = molar vapor and liquid flow

[m] [–] [mol m–2 s–1]

HETP data on several common packing are given in Fig. 4.15 as a function of vapor rate. It can be seen that smaller packings improve the HETP, however, this advantage is counteracted by a similar increase in pressure drop. Conventional

Fig. 4.15 HETP, height equivalent of a theoretical plate for commercial available packings (adapted from [13]).

102

4 General design of gas/liquid contactors

types of packings, such as Raschig and saddle type rings, exhibit a range of HETP-values as a function of vapor rate, whereas newer types, such as Mini rings and Pall rings, give a fairly constant value over a wide range of vapor rates. 4.4.4.2 Transfer units The other approach is based on a hydrodynamic model of mass transfer in a packed column. A packed G/L contactor can be seen at as a column where liquid and vapor flow countercurrently. Fig. 4.16 shows the essentials to formulate the differential mass balance. The packed part of the column contains Ah unit surface area per unit length of column height. Applying the rate expression given in Eq. 4.6, the mass balance over a thin slice of height Δh becomes: V ⋅ y = V ⋅ (y + Δy) + k OV ⋅ ρV ⋅ (y − y ∗ ) ⋅ Δh ⋅ Ah

(4.31)

Integrating from y = yin at h = 0 to y = yout at h = Hcolumn and rearranging leads to yout

Hcolumn =

Qv k OV ⋅ Ah

∫

yin

dy ∗

y −y

= Htu ⋅ NOV

(4.32)

The first term at the right hand side of this equation is called the height of a transfer unit, Htu , calculated from volume flow rate, overall mass transfer coefficient and surface area provided by the packing, similar to Eq. 4.14. The integral is defined in Eq. 4.12 as the overall number of transfer units, NOV . Where the

Fig. 4.16 Mass transfer in an idealized packed bed absorber.

4.4 Packed columns

103

height of a transfer unit reflects the ratio of convective transport and absorption rate, the number of transfer units determines the axial concentration gradient over the column: yout

Htu ≡

Qv k OV ⋅ A h

and

NOV ≡

∫

yin

dy ∗

y −y

(4.33)

In contrast with y* in the analogous equation for a sieve tray (Eq. 4.12), y* in the mass balance for a packed column is not a constant but a function of the local liquid composition x. The integral can be solved in case of linear operating (Eq. 3.17 for absorption) and equilibrium lines (Eq. 3.13): y − yout =

L

(x − x in)

V

(4.34)

and y* = K ⋅ x

(4.35)

Elimination of x and y* from Eqs. 4.33–4.35 and putting the separation factor S = Sabs = L/KV (in case of absorption S equals the absorption factor A) gives the following expression for NOV :

NOV

1 (yout − yin) + yin − Kx in S S ln = S−1 yout − Kx in =

S S−1

ln

yin − Kx in

1−

[yout − Kx in (

1 S)

+

1 S]

(4.36)

With Eqs. 4.32 and 4.36 the column height can be calculated. On the other hand, the number of transfer units, NOV, is connected to the number of theoretical stages, Nts , as calculated with the Kremser equation through the separation factor S. Comparing Eqs. 3.28 (for the case of absorption S = A) and 4.36 gives: NOV = Nts ⋅

S ⋅ ln S S−1

(4.37)

with separation factor S = Sabs = A = L/KV in case of absorption or S = Sstr = KV/L in case of stripping. Eq. 4.37 defines the relation between the number of transfer units and the number of theoretical stages. As an alternative to Eq. 4.29, the column height may also be calculated from: Hcolumn = Htu ⋅ NOV

(4.38)

104

4 General design of gas/liquid contactors

4.4.5 Minimum column diameter In contrast to tray columns, Fig. 4.17 illustrates that packed columns require a minimum liquid load to ensure sufficient mass transfer. Below this minimum value, only a very small part of the packing surface is wetted and liquid and gas are no longer in intimate contact. This results in a considerable drop of separation efficiency. The lower capacity limit of liquid load depends on the type of the packing, the quality of the packing supplements and the physical properties of the liquid. The diameter of a packed bed column is determined by the required maximum capacity. Finding the maximum allowable vapor rate requires a basic understanding of how flooding and pressure drop are related to operating conditions and packing properties. The feasible operating region of packed columns differs considerably from tray columns and is limited by flooding and wetting. As illustrated by Fig. 4.17, flooding sets the upper capacity limit. At the flood point the pressure drop of the gas flow through the bed increases so drastically that the liquid is no longer able to flow downward against the gas flow. Hence, the countercurrent flow of gas and liquid breaks down and the separation efficiency decreases dramatically. Flooding correlations for packed beds are available in plots similar to those for tray columns. The abscissa scale term is the same flow parameter (a measure of the ratio of kinetic energies of liquid and vapor) used for trays. The dependent parameter should contain parameters characteristic for the actual flow rate,

Fig. 4.17 Operating region of a packed column.

4.4 Packed columns

Fig. 4.18

105

Flooding properties of a packed column

vV, flood , and properties of the packing, such as ε/a, the ratio of volume available for flow and the wetted surface area of the packing, where ε is the fractional void volume of the dry packing and a the packing surface per unit column volume. Converting the actual flow rate, vV, flood , to the superficial10 (or empty tube) vapor rate uV, flood by uV, flood = ε ⋅ vV, flood then a relevant dimensionless parameter a ρV ⋅ . The ratio a/ε3 is a function of the packing appears to be uV, flood 3 √ε g ρL properties only and is usually referred to as the packing factor Fp [m–1]. In Fig. 4.18 the flooding properties of packed beds are given in a plot of Fp ρV ρL QL uV, flood ⋅ as a function of the flow parameter . The ordinate in QV √ρV √ g ρL this figure and in that of Fig. 4.10 for sieve trays are similar, 1/Fp taking the place of Hspacing . The packing factor Fp for a certain type of random packing is either supplied by the manufacturer or estimated from ε and a. Like the flooding diagram of sieve trays, the latter diagram is also subject to some restrictions. The flooding line in Fig. 4.18 is only applicable to – non-foaming systems, – flooding defined as a pressure drop of 0.01 bar m–1, and – systems with viscosity equal to that of aqueous solutions 10

The total flow rate divided by the cross sectional area of the column.

106

4 General design of gas/liquid contactors

For systems with a viscosity η different from water, the maximum vapor rate calculated from the ordinate value should be multiplied by (ηwater /η)0.1. Once QV , the flow parameter and the packing factor are known, the flooding velocity of the vapor can be calculated from the value read at the ordinate. For design purposes the maximum allowable vapor rate is usually taken at some 70 % of the flooding velocity uV, flood. Once this value has been fixed, minimum column cross sectional area and minimum column diameter can be calculated in a way similar to the dimensions of a tray column as in Eq. 4.26:

Apacking, min =

QV 0.70 ⋅ uV, flood

and

Dmin =

4

√π

⋅ Apacking, min

(4.39)

Packing factors of, e.g., Raschig rings range from 100 to several hundreds reciprocal meters, depending on size and material. Structured packings show much lower values; flooding velocities in structured packings can be two to three times higher, resulting in 40–70 % smaller column diameters. 4.4.6 Pressure drop The column diameter as estimated in the previous section is the minimum column diameter as far as flooding is considered. In this estimation constant column pressure is assumed. This assumption should be checked. The pressure drop ΔP over a packed column can be calculated from the wellknown Ergun equation ΔP L where L = Dp = ε = ρ¯ = Re η

11

=

=

Go2 Dp ⋅ ρ¯

⋅

1−ε ε

3

⋅ 1.75 +

(

150

length of packed bed11 effective diameter of packing material bed porosity average gas phase density in column ρ vo Dp

(1 − ε) η = viscosity

(4.40)

Re ) [m] [m] [–] [kg m–3] [–]

[Pa.s], [kg m–1 s–1]

In case of distillation, L refers to the length of either the rectifying or stripping section.

4.5 Criteria for column selection

v0 G0

107

[m s–1] [kg s–1 m–2]

= superficial velocity = ρvo , mass flux

It should be noted that, while ρ and v0 vary across the column, the product ρvo is constant. This Ergun equation has been applied successfully to gases by using the average density of the gas, provided the pressure drop over the column is sufficiently small. For Re > 1000 the Ergun equation reduces to the Burke-Plummer equation for turbulent flow. For small values of the Reynolds number (Re < 10) the term 1.75 is small compared to 150/Re and Eq. 4.40 transforms into the Blake-Kozeny equation for laminar flow (which is based on Darcyʼs law: ΔP/L = v0 η / permeability) Estimation of the total pressure drop starts with the assumption that at the conditions of the minimum column diameter the pressure drop is reasonable small and that the ideal gas law is applicable. Then, the average density follows from ρ¯ =

¯ M

Ptot −

ΔP

RT(

(4.41)

2 )

¯ being the molecular weight of the gas phase mixture. The minimum column M diameter defines the superficial velocity v0 at the inlet of the column. With the known value of v0 at that position G0 and Re are calculated G0 =

¯ ⋅ Ptot M RT

⋅ v0

and

Re =

G0 ⋅ Dp (1 − ε)η

(4.42)

which are constant throughout the column. Elimination of ρ¯, G0 and Re by combining Eqs. 4.40–4.42 results in a single equation with just one unknown: ΔP. A convenient way to solve the complete set of three equations is by iteration: a) assume a pressure drop of say 10 % of the total pressure, b) calculate ρ¯ (Eq. 4.41) and G0 and Re (Eqs. 4.42) c) calculate ΔP from Eq. 4.40 and compare the result with the value assumed in a. d) if ΔP from Eq. 4.40 is larger than the value used in a, the pressure drop is to large, Dmin should be increased to decrease v0 and ΔP. e) otherwise, to refine the result, the calculation can be repeated starting at b with the smaller value of ΔP.

4.5 Criteria for column selection Proper choice of equipment is very important for effective and economical distillation. Tray columns are generally employed in large diameter (> 1 m) towers. The gas load must be kept within a relatively narrow range, only valve trays allow greater operational flexibility. The liquid load can be varied over a wide range

108

4 General design of gas/liquid contactors

even to very low liquid loads. For vacuum operation their relatively high pressure drop (typically 7 mbar per equilibrium stage) is a disadvantage. Tray columns have a relatively high liquid holdup compensating for fluctuations in feed compositions. However, the resulting high liquid residence time in the column may lead to the decomposition of thermally unstable substances. Tray columns are also relatively insensitive to impurities in the liquid. Packed columns are used almost exclusively in small diameter (< 0.7 m) towers. Only structured packings permit the use of packed columns with very large diameters. They are extremely flexible with regard to gas load, but require a minimum liquid load for packing wetting. Structured packings with an extremely small pressure loss, typically under 0.5 mbar per equilibrium stage, are a tremendous advantage in vacuum operation. The danger of decomposition of thermally unstable substances is also less in packed columns because of very small liquid holdup. Nomenclature a A Ah d, D DAB EMV EO Fp g G0 H HETP, HETS, Hets Hspacing HTU, Htu k K L m NOV Ns Nts Q R S u v x, y V V z ΦA ϕ ε

surface area per unit column volume, page 105 area (interfacial, column cross section) interfacial area per unit column height, Eq. 4.14 diameter (of packing rings, column or particles) coefficient of diffusion of A in a mixture of A and B Murphree or plate efficiency, Eq. 4.10 overall efficiency, Eq. 4.9 packing factor, page 105 acceleration constant due to gravity mass flux, Eq. 4.40 height height equivalent of a theoretical stage, Eq. 4.29 distance between two adjacent plates height of a transfer unit, Eq. 4.14 mass transfer coefficient equilibrium constant liquid flow volumetric distribution coefficient, Eqs. 4.6, 4.7 overall number of gas-phase transfer units, Eq. 4.12 number of real stages number of theoretical stages volume flow gas constant separation factor, S = A = L / KV for absorption S = KV / L for stripping superficial velocity, page 105 interstitial velocity, page 105 mole fraction vapor flow total system volume (liquid + vapor), in Eq. 4.2 distance from interface molar transfer rate of component A hole area over active or bubble area on a sieve tray fractional void volume

[m2m–3] [m2] [m2m–1] [m] [m2s–1] [–] [–] [m–1] [m s–2] [kg m s–1] [m] [–] [m] [m] [m s–1] [–] [mol s–1] [–] [–] [–] [–] [m3s–1] [J mol–1K–1] [–] [m s–1] [m s–1] [–] [mol s–1] [m3] [m] [mol s–1] [–] [–]

Exercises

δ ρ ρ γ

film thickness density, Figs. 4.10 and 4.18 molar density, Eq. 4.1 surface tension

109

[m] [kg m–3] [mol m–3] [N m–1]

Indices A, B abs C i L LV

components absorption column interface liquid liquid-vapor interface

OL , OV overall based on resistance in liquid and vapor phase, respectively i interface str stripping V vapor z at distance z from interface

Exercises 1. An aqueous droplet with radius R is surrounded by a stagnant, inert vapor phase containing a low concentration of ammonia. Show that y | r − y | r = R , the difference in ammonia concentration in the vapor phase and at the LV-interface, is proportional to 1/R – 1/r, where r is the distance to the centre of the droplet. Assume that the resistance against mass transfer from the vapor to the liquid is mainly situated in the vapor phase. 2. Derive the rate expression in Eq. 4.6, starting from the stationary state balance given in Eq. 4.3. 3. A tray column is to be designed to reduce the water content in natural gas (Mmethane = 0.016 kg mol–1) from 0.10 to 0.02 mole% by absorption of water at 70 °C and 40 bar in diethylene glycol (MDEG = 0.106 kg mol–1, ρDEG = 1100 kg m–3) containing 2.0 mol% H2O. The equilibrium constant K = 5.0 · 10–3 at this temperature. Preliminary calculations (see exercise 3.5) resulted in a slope of the operating line L / V = 0.010 and Nts = 2.25. The column should have the capacity to treat QV = 3 m3 gas per unit time at this condition (40 bar, 70 °C). Lab scale experiments showed that the overall mass transfer coefficient kOV = 3 · 10–3 m s–1 at this gas load. The gas phase is assumed to obey the ideal gas law. Hspacing = 0.5 m, g = 9.81 m s–2. Calculate: a. the minimum column diameter b. the plate efficiency c. the height of the column 4. An aqueous waste stream of 0.015 m3s–1, saturated with benzene, should be purified by reducing the benzene content by at least 99.9 %. This is possible by stripping with (pure) air. To this purpose an existing tray column with 5 trays is available. The effective surface area of a single tray amounts to 1.77 m2. Laboratory experiments at 1 bar and 294 K show that the overall mass transfer coefficient based on the gas phase, kOV = 0.0080 m s–1. The stripping process is carried out isothermically at the same conditions with an airflow12 of 0.932 m3STPs–1. The distribution coefficient Kbenzene = 152, and the satu12

1 m3 STP means 1 m3 at standard temperature (0 °C) and standard pressure (1 atm).

110

4 General design of gas/liquid contactors

ration pressure of benzene is 0.104 bar. At this air flow the interfacial surface area is 50 m2 per m2 tray area. Mair = 0.029 kg mol–1, Mwater = 0.018 kg mol–1, Mbenzene = 0.078 kg mol–1, g = 9.8 ms–2, R = 8.31 J mol–1 K–1, ρwater = 1000 kg m–3, 1 atm = 1.01325 bar. a. How many theoretical stages are required to reduce the benzene content by 99.9 %? b. Calculate the plate efficiency EMV. c. Show by calculation that at the given flow rate with this column the required reduction can be obtained. d. Calculate the mole fraction of benzene in the effluent. e. Comment on the chosen value of the airflow. f. How should the exiting gas flow be treated to avoid dumping of benzene? 5. An air-ammonia mixture, containing 5 mol% NH3 at a total flow rate of 5 mol s–1, is scrubbed in a packed column by a countercurrent flow of 0.5 kg water s–1. At 20 °C and 1 bar 90 % of the ammonia is absorbed. Mair = 0.029 kg mol–1, Mammonia = 0.017 kg mol–1, Mwater = 0.018 kg mol–1, the density of water may be taken as 1000 kg m3. Calculate the flooding velocity for two different packing materials: a. 25 mm ceramic Raschig rings with a packing factor of 540 m–1, and b. 25 mm metal Hiflow rings with a packing factor of 125 m–1. 6. To study the rate of absorption of SO2 in water, a laboratory scale column packed with plastic Hiflow rings is used. The diameter of the rings is 15 mm. The bed has a porosity εbed = 0.90 and a surface area per unit bed volume a = 200 m2 m–3. The height of the packing is 2.0 m, the internal diameter of the column 0.33 m. Air containing 2.0 mol% SO2 is fed to the column at a rate of V = 2.25 mol s–1, clean water is fed to the top of the column. The absorption of SO2 in water is studied in counter-current operation at 1 bar and 285 K. At this temperature the distribution coefficient KSO2 = 32. At a liquid flow rate L = 65 mol s–1 the SO2-content in the effluent is reduced to 0.50 mol%. Mlucht = 0.029 kg mol–1, Mwater = 0.018 kg mol–1, g = 9.81 m s–2, R = 8.31 J mol–1 K–1, ρwater = 1000 kg m–3. a. How much larger is the chosen value of L compared to the theoretical minimum value, Lmin? b. How many transfer units NOV characterize this absorption process at the given process conditions? c. Calculate the overall mass transfer coefficient based on the gas phase, kOV. c. At what (superficial) velocity of the gas feed flooding will start? d. Calculate the concentration of SO2 in the gaseous effluent in case the air feed is increased to 75 % of the flooding velocity. Temperature, pressure and liquid flow rate remain unchanged.

5 Liquid-liquid extraction

5.1 Liquid-liquid extraction 5.1.1 Introduction Since the introduction of industrial liquid-liquid extraction processes, a large number of applications have been proposed and developed. An overview of some important industrial example applications is presented in Table 5.1. The first and until now largest-volume industrial application of solvent extraction is in the petrochemical industry. Extraction processes are well suited to the petroleum industry because of the need to separate heat-sensitive liquid feeds according to chemical type (e.g. aliphatic, aromatic, naphthenic) rather than by molecular weight or vapor pressure. Other major applications include the purification of antibiotics and the recovery of vegetable oils from natural substrates. In metals processing the recovery of metals such as copper from acidic leach liquors and the refining of uranium, plutonium and other radioactive isotopes from spent fuel elements. Recently extraction is gaining increasing importance as a separation technique in biotechnology. Liquid-liquid extraction is based on the partial miscibility of liquids and used to separate a dissolved component from its solvent by transfer to a second solvent. The principle of liquid-liquid extraction, and some of the special terminology is

Table 5.1

Industrial liquid-liquid extraction processes.

Solute

Carrier

Solvents

Acetic Acid Aromatics

Water Paraffins

Caprolactam Benzoic Acid Formaldehyde Phenol Penicillin Vanilla Vitamins A, D Vitamin E Copper Uranium

Aqueous Ammonium Sulphate Water Water Water Broth Oxidized liquors Fish liver oils Vegetable oils Acidic leach liquors Acidic leach liquors

Ethyl acetate, Isopropyl acetate Diethylene glycol, Furfural, Sulpholane, NMP, DMSO Benzene, Toluene, Chloroform Benzene Isopropyl ether Benzene Butyl acetate Toluene Liquid propane Liquid propane Chelating agents in kerosene Tertiary amines in kerosene

112

5 Liquid-liquid extraction

Fig. 5.1 Principle of a single extraction contacting stage.

illustrated in Fig. 5.1. In the simplest case the feed solution consists of the carrier solvent containing the desired solute. This liquid feed is contacted with a solvent, which is immiscible or only partly miscible with the liquid feed. In general the solvent has a higher affinity for the solute than the carrier and the solute is extracted into the solvent phase. For efficient contact a large interfacial area must be created across which the solute can transfer until equilibrium is closely approached. This is achieved by bringing the feed mixture and the solvent into intimate contact. The resulting solute loaded solvent phase is called the extract, while the other liquid phase is designated the raffinate. When equilibrium is reached, the stage is defined as an ideal or theoretical stage and the equilibrium conditions can be expressed in terms of the extraction factor E for the solute: E=

amount of solute in extract phase amount of solute in raffinate phase

=K

Extract flow Raffinate flow

(5.1)

where K is the distribution coefficient of the solute between the extract and the raffinate. The larger the value of the extraction factor E, the greater the extent to which the solute it extracted. Large values of E result from large values of the distribution coefficient, K, or large solvent to carrier ratios. After extraction at least one distillation column (or other separation process) is required to separate the solvent from the extract and recycle the solvent. If the solvent is partially miscible with the feed, a second separation process (normally distillation) is required to recover solvent from raffinate. Fig. 5.2 illustrates such an extraction system where two distillation columns are needed. This example is based on a high-boiling solvent, which is recovered as the bottom product of the distillation column. A low-boiling solvent would be recovered as the top product. Solvent recovery is an important factor in the economics of industrial extraction processes. Especially if the solvent and solute have close boiling points, the distillation column may require many trays and a high reflux ratio, resulting in a costly process. The main disadvantage of extraction is that the necessity of a solvent increases the complexity and thereby costs of the process. Situations where extraction is preferred to distillation and should be considered include the following application areas:

5.1 Liquid-liquid extraction

113

Fig. 5.2 Schematic of a principle liquid-liquid extraction process.

• Dissolved or complexed inorganic substances in organic or aqueous solutions • Removal or recovery of components present in small concentrations • When a high-boiling component is present in relatively small quantities in a waste steam • Recovery of heat-sensitive materials and low to moderate processing temperatures are needed • Separation of a mixture according to chemical type rather than relative volatility • Separation of close-melting or close-boiling liquids, when solubility differences can be exploited • Mixtures that form azeotropes or exhibit low relative volatilities and distillation cannot be used 5.1.2 Liquid-liquid equilibria Addition of a solvent to a binary mixture of the solute in a carrier can lead to the formation of several mixture types. The simplest ternary system is the Type I system for one immiscible-liquid pair shown in Fig. 5.3. In such a system, the carrier and the solvent are essentially immiscible, while the solute is miscible with the carrier as well as the solvent. Such ternary systems are commonly represented in a triangular diagram, showing the two-phase envelope that encloses the

114

5 Liquid-liquid extraction

Fig. 5.3 Triangular diagram for a Type 1 system with the tie-lines represented by the dashed lines.

region of overall compositions in which two phases exist in equilibrium. The equilibrium compositions of the extract and raffinate phases are connected by tielines that can be used to determine distribution coefficients and selectivities. For each component the distribution coefficient is given by the ratio of the concentrations in the two phases as: Ki =

xi, E

(5.2)

xi, R

where xi,E is the mole fraction of component i in the extract phase and xi,R is the mole fraction of component i in the raffinate phase. The selectivity is, equivalent to the relative volatility in distillation, defined as the ratio of K values:

β12 =

K1 K2

x1 (x2) =

extract

x1 (x2)

(5.3)

raffinate

At equilibrium, thermodynamics requires the activity of each component to be the same in the two liquid phases. However, due to the strong thermodynamic non-ideal behavior of the two liquid phases, the large difference in activity coefficients causes significant differences in equilibrium compositions: γi, E xi. E = γi, R xi. R

(5.4)

5.1 Liquid-liquid extraction

115

Replacing the composition ratio with the ratio of activity coefficients gives for the selectivity: γ2 (γ1) β12 =

extract

γ2 (γ1)

(5.5)

raffinate

The required activity coefficients or distribution coefficients are generally calculated from thermodynamic models such as NRTL, Uniquac or predicted from group contribution methods like UNIFAC. In Fig. 5.3 it can be seen that as the concentration of the solute is increased, the tie-lines become shorter because of the increased mutual miscibility of the two phases. At the plait point the raffinatephase and extract-phase boundary curves intersect and the selectivity becomes equal to one. An important additional use of the triangular diagram is the graphical solution of material balance problems, such as the calculation of the relative amounts of equilibrium phase obtained from a given overall mixture composition. Liquid-liquid equilibria having more than three components cannot be represented on a two-dimensional diagram. 5.1.3 Solvent selection The key to an effective extraction process is the selection of the right solvent. A solvent or extractant may be a pure chemical compound or a mixture of chemical compounds. Proposing a solvent requires the knowledge of the physical and chemical properties of the solvent, solute, carrier and other constituents of the extraction system. While pure component properties are easily found in the literature, mixture physical properties are available in the literature for very few systems. The following criteria are important to consider during the selection of a solvent: • Distribution coefficient – A high value of distribution coefficient indicates high affinity of the solvent for the solute and permits lower solvent/feed ratios • Selectivity − A high value of the selectivity reduces the required number of equilibrium stages. If the feed is a complex mixture where multiple components need to be extracted from, selectivities between groups of components become important. • Density – Higher density differences between extract and raffinate phases permits higher capacities in extraction devices using gravity for phase separation. The density difference must at least be large enough to ease the settling of the liquid phases. • Viscosity – High viscosities lead to difficulties in pumping, dispersion, and reduce the rate of mass transfer. Low viscosities benefit rapid settling, capacity

116

•

• •

•

•

•

• •

5 Liquid-liquid extraction

and formation of smaller droplets, which have larger specific surface areas (see interfacial surface area ALV in Eq. 4.2) Solvent recoverability – Recovery of the solvent should be easy. A solvent that boils much higher than the solute generally leads to better results, though solvents boiling lower than the solute are also used commercially. Solubility of solvent – Mutual solubilities of carrier and solvent should be low to avoid an additional separation step for recovering solvent from the raffinate. Interfacial tension – Low interfacial tension facilitates the phase dispersion but may require large volumes for phase separation due to slow coalescence. High interfacial tension permits a rapid settling due to an easier coalescence, allowing higher capacities. Too low an interfacial tension leads to emulsification. Availability and cost – Solvent cost may represent a large initial expense for charging the system, as well as a heavy continuing expense for replacing solvent losses. Therefore, one should make sure that the solvent of interest is commercially available and relatively inexpensive. Toxicity, compatibility and flammability – These criteria are important occupational health and safety considerations a suitable solvent has to meet. Especially for food and pharmaceutical products only nontoxic solvents will be taken into consideration. In general, any hazard associated with the solvent will require extra safety measures and increases costs. Thermal and chemical stability – It is important that the solvent should be thermally and chemically stable because it is recycled. In particular the solvent should resist breakdown during recovery in for example a distillation column. Corrosivity – Corrosive solvents can lead to increased equipment cost but might also require expensive pre- and post-treatment of streams. Environmental impact – The solvent should not only be compatible with the process, but also with the environment (minimal losses due to evaporation, solubility and entrainment)

In addition to being nontoxic, inexpensive and easily recoverable, a good solvent should be relatively immiscible with feed components other than the solute and have a different density from the feed to facilitate phase separation. Also, it must have a very high affinity for the solute, from which it should be easily separated by distillation, crystallization or other means. Obviously no solvent will be best from all of these viewpoints, and the selection of a desirable solvent involves compromises between the various criteria. In this selection process some or the criteria are essential for the separation while others are desirable properties that will improve the separation and/or make it more economical. The most important compromise is always between solute solubility and selectivity that usually behave exactly opposite. When a high selectivity is obtained, solubilities are usually low and the other way around. In most cases, the selectivity is the most important parameter, since this determines whether a certain separation can be accomplished.

5.2 Extraction schemes

117

5.2 Extraction schemes In the simplest extraction scheme a feed and a solvent are contacted in a single equilibrium stage as shown in Fig. 5.1. The main disadvantage of single stage contacting is that residual solute is left behind in the carrier. Therefore often a series of contacting stages is arranged in a cascade to accomplish a separation that cannot be achieved in a single stage and/or reduce the required amount of the mass-separating agent. Three frequently encountered extraction cascades are the cocurrent, crosscurrent and countercurrent arrangements (Fig. 5.4). In this paragraph we will explore their difference in extraction efficiency. All examples in this section refer to dilute systems where the distribution coefficient K can be considered constant.

Fig. 5.4 Schematic of (a) cocurrent; (b) crosscurrent and (c) countercurrent extraction.

5.2.1 Single equilibrium stage Consider the case that the solvent and the carrier are completely immiscible, the solvent contains no solute and solute concentrations in the extract and raffinate remain low (< 1 %). Under these conditions the resulting extract flow will approximately equal the solvent flow, E ≈ S, and the raffinate flow approximately equal the feed flow, R ≈ F. Thus the material balance over the solute in a single equilibrium stage becomes: xF F = xE S + xR F

(5.6)

and the distribution of solute at equilibrium is given by xE = KxR

(5.7)

118

5 Liquid-liquid extraction

where K is the distribution coefficient defined in terms of mass or mole ratios. Elimination of xE gives: xR =

xF F F + KS

=

xF KS 1+ F

(5.8)

It is convenient now to introduce the previously defined extraction factor1, E, for the solute: E=

K E1 R1

=

KS F

(5.1)

The fraction of solute that is not extracted is now given by: xR xF

=

1 1+E

(5.9)

where it is clear that the larger the extraction factor, the greater the extent to which the solute is extracted. Large values of E result from large values of the distribution coefficient, K, or large ratios of solvent to carrier. 5.2.2 Cocurrent cascade If additional stages are added in the cocurrent arrangement in Fig. 5.4a, the equation for the first stage is that of a single stage: x1 xF

=

1 1+E

(5.10)

Since y1 is in equilibrium with x1 (y1 = Kx1) this can be rewritten as: y1 xF

=

K 1+E

(5.11)

For the second stage, a material balance for the solute gives: x1 F + y1 S = x 2 F + y 2 S

(5.12)

with: K=

1

y2 x2

(5.13)

Note that E = extraction factor, not to be confused with E1, E2, … the extract flow rate from stage 1, 2, … as in Fig. 5.4.

5.2 Extraction schemes

119

Combining (5.11) with (5.12), (5.10) and y2 = Kx2 to eliminate respectively y1, x1, and y2 gives: x2 xF

=

1

=

1+E

xN xF

(5.14)

Comparison of (5.14) with (5.10) shows that xN = x1, meaning that no additional extraction takes place in the second stage. This is as expected because in the derivation we assumed that the two streams leaving the first stage are at equilibrium and when they are recontacted in stage 2, no additional net mass transfer of solute occurs. Accordingly, a cocurrent cascade only provides increased residence time that is not necessary when the stages are ideal because equilibrium is reached between the streams after the first stage. 5.2.3 Crosscurrent For the crosscurrent cascade shown in Fig. 5.4b, the feed progresses through each stage, starting with stage 1 and finishing with stage N. When the total amount of solvent S is distributed in equal portions S/N over N stages, the extraction factor for each stage becomes E/N. Application of Eq. 5.10 by replacing E by E/N gives the concentration ratios in each of the N stages: x1 xF

1

=

1+

( x2 x1

=

E N)

1 E 1+ N) ( : :

xN xN−1

(5.15)

1

=

1+

(

E N)

Combining the equations in (5.15) to eliminate all intermediate interstage variables, xN , the final raffinate mass ratio is given by: xN xF

=

1 E N 1+ N) (

(5.16)

120

5 Liquid-liquid extraction

Thus, unlike the cocurrent cascade, the value of xN decreases in each successive stage and the distribution of fresh solvent over multiple stages gives an improvement compared to the separation in a single stage for a given solvent to feed ratio. For an infinite number of equilibrium stages, (5.16) becomes: x∞ xF

=

1

(5.17)

exp (E )

Thus, even for an infinite number of stages, xN , cannot be reduced to zero. 5.2.4 Countercurrent The best compromise between the objectives of high extract concentration and a high degree of extraction of the solute is the countercurrent arrangement. In the countercurrent arrangement in Fig. 5.4c, the feed entering stage 1 is brought into contact with a solute-rich stream, which has already passed through the other stages, while the raffinate leaving the last stage has been in contact with fresh solvent. For a three-stage system, the material balances and equilibrium equations for the solute in each stage are as follows: Stage 1:

xF F + y 2 S = x1 F + y1 S

(5.18)

Stage 2:

x1 F + y 3 S = x 2 F + y 2 S

(5.19)

Stage 3:

x 2 F + yS S = x 3 F + y 3 S

(5.20)

Equilibrium

K=

y1 x1

=

y2 x2

=

y3

(5.21)

x3

Solving this set of equations the ratio between raffinate and feed composition, x3 /xF = xR /xF, is obtained by elimination of y1, y2 and y3 from Eqs. 5.18–5.21. Taking the mole fraction in the entering solvent yS = 0, introduction of the extraction factor E = K · S/F and subsequent elimination of x1 and x2 from the resulting set of three equations gives: xR

=

xF

1 1 + E + E2 + E3

=

E−1 E4 − 1

(5.22)

If the number of countercurrent stages is extended from n = 3 to n = N, this becomes: xR xF

=

1

=

N

∑E

n=0

n

E−1 E N+1 − 1

(5.23a)

5.3 Design of countercurrent extractions

121

As with the crosscurrent arrangement, the value of xN = xR decreases in each successive stage. For the countercurrent arrangement this decrease is greater than for the crosscurrent arrangement. Therefore, the countercurrent cascade is the most efficient of the three linear cascades and normally preferred for commercial scale operations because of economic advantages. Expressing N in Eq. 5.23a as a function of xR /xF gives the Kremser-equation for extraction, similar as that for absorption (Eq. 3.28) and that for stripping (Eq. 3.33). The complement of xR /xF is the fraction extracted: 1−

xR

=1−

xF

E−1 E N+1 − 1

=

E N+1 − E E N+1 − 1

(5.23b)

similar to the fraction absorbed (Eq. 3.29) and the fraction stripped (Eq. 3.34).

5.3 Design of countercurrent extractions 5.3.1 Graphical McCabe-Thiele method for immiscible systems The McCabe-Thiele analysis for immiscible extraction is very similar to the analysis for absorption and stripping discussed in Chapter 3. In order to use a McCabeThiele type of analysis we must be able to plot a single equilibrium curve, have the energy balances automatically satisfied, and determine the operating line. The energy balances are automatically satisfied when it is assumed that the heat of mixing is neglible and the system is isothermal as well as isobaric. The operating line will be straight and the solvent and diluent mass balances are automatically satisfied when the carrier and solvent are assumed totally immiscible. In that case the solute free carrier flow rate, F ′, and solute free solvent flow rate, S ′, remain constant. F ′ = solute free carrier flow rate, kg/s = constant

(5.24)

S ′ = solute free solvent flow rate, kg/s = constant

(5.25)

Note again that these are flow rates of carrier and solvent only and do not include the total raffinate and extract streams. Eqs. 5.24–25 are the carrier and solvent mass balances that are automatically satisfied when the phases are immiscible. When the carrier and solvent are immiscible, weight ratio units are related to weight fractions as: X=

x 1−x

and

Y=

y 1−y

(5.26)

where X is kg solute/kg carrier and Y is kg solute/kg solvent. The operating equation is derived from the mass balance envelope shown in Fig. 5.5. In weight ratio units the steady-state mass balance is:

122

5 Liquid-liquid extraction

Fig. 5.5 Mass balance envelope of a countercurrent extractor.

F ′X n + S ′Y1 = S ′Yn+1 + F ′X 0

(5.27)

Solving for Yn+1 we obtain the operating equation: Yn+1 =

F′

(S ′)

Xn + Y1 −

[

F′

X0

(S ′) ]

(5.28)

When plotted on a McCabe-Thiele diagram of Y versus X, this is a straight line with slope F ′ /S ′ and Y intercept Y1 – (F ′ /S ′ ) X0 , as illustrated in Fig. 5.6. Note that since F ′ and S ′ are constant, the operating line is straight. For the usual design problem, F ′ /S ′ will be known as well are X0 , XN and YN+1. Since XN and YN+1 are the concentrations of passing streams, they represent the coordinates of a point on the operating line. Equilibrium data for dilute extraction are usually represented as a distribution ratio: KB =

yB xB

(5.29)

in weight fractions or mole fractions. For very dilute systems the distribution ratio will be constant, while at higher concentrations it often becomes a function of concentration. The distribution coefficient is nearly always temperature and pH dependent. The McCabe-Thiele diagram shown in Fig. 5.6 is very similar to McCabeThiele diagrams for stripping. This is because the processes are analogous in that

5.3 Design of countercurrent extractions

123

Fig. 5.6 McCabe-Thiele diagram for dilute extraction.

both contact two phases and solute is transferred from the X phase to the Y phase. In both analyses we defined constant flows in each phase and used ratio units. The analogy breaks down when we consider stage efficiencies and sizing of the column diameter, since mass transfer characteristics and flow hydrodynamics are very different for extraction and stripping. If the system is very dilute (< 1 mole or weight percent solute), total flows will be constant and the McCabe-Thiele diagram can be directly plotted in fractions. Since adding large amounts of solute often makes the phases partially miscible, the assumption of complete immiscibility is often valid only for dilute systems. Similar to absorption and stripping a minimum solvent flow rate exists (see Eq. 3.7 and Fig. 3.6) for each feed flow rate, solute concentration, extent of extraction and operating temperature, that corresponds to an infinite number of countercurrent equilibrium contacts between the extract and raffinate phases. As a result a tradeoff exists in every design problem between the number of equilibrium stages and the solvent flow rate at rates greater than the minimum value. ′ is obtained from a mass balance This minimum solute free solvent flow rate S min over the whole extractor with an infinite number of stages. An overall mass balance over the column illustrated by Fig. 5.5 gives the following result: S ′Yin + F ′Xin = S ′Yout + F ′Xout

(5.30)

The assumption of an infinite number of stages (in other words: a column of infinite length) means that the outgoing extract is in equilibrium with incoming

124

5 Liquid-liquid extraction

feed (Yout = K · Xin or Y1 = K · X0, where α in Eq. 3.13 is replaced by equilibrium constant K) allows the calculation of the solute weight ratio Yout in the extract which is in equilibrium with the solute weight ratio Xin in the feed. Reorganizing Eq. 5.30 then gives the minimum solute free solvent flow (compare with Eq. 3.7): ′ Smin = F′

Xin − Xout

Yout, max = K · Xin

with

Yout, max − Yin

(5.31)

′ 2 must be selected For economical operation a solvent flow rate greater than S min for the extraction to be conducted in a finite number of stages. A reasonable ′ . starting value for S ′ is usually 1.5 · S min

5.3.2 Analytical Kremser method If one additional assumption can be made, the Kremser equation can also be used for dilute extraction. Obviously, this assumption is that the distribution coefficient is constant. The dilute extraction model now satisfies all the assumptions used to derive the Kremser equations in Chapter 3, so these are applicable: Fraction of a solute, i, extracted =

xin − xout xin

yin − K

=

E N+1 − E E N+1 − 1

(5.32)

In the case that the solvents are substantially immiscible and a constant distribution coefficient the required number of theoretical stages, Nts , for countercurrent contact can be calculated directly from the Kremser equation: xin − ln

Nts =

[ (x

out

−

yin K yin K

)

1 −

(

1 1 + E) E

]

ln E

(5.33)

5.3.3 Graphical method for partial miscible systems For higher solute concentration containing partially miscible ternary systems, stage wise extraction calculations are conveniently carried out with ternary equilibrium diagrams. In this section, triangular diagrams are used for the development and illustration of a graphical calculation procedure. Phase equilibrium may 2

For the determination of Smin in partial miscible systems see Fig. 5.10.

5.3 Design of countercurrent extractions

125

Fig. 5.7 Location of product points in ternary diagram.

be represented as illustrated in Fig. 5.7 on a triangular diagram3, where the dashed lines are the tie lines that connect the equilibrium phases of the equilibrium curve. On the equilibrium curve all extract compositions lie to the left of the plait point, while all raffinate compositions lie to the right. The problem posed is to determine the number of stages given the flow rates and compositions of the feed and solvent, and the desired raffinate composition. For this we consider a countercurrent, contactor with N stages for liquid-liquid extraction of a ternary system operating under isothermal, continuous, steady state flow conditions, as shown in Fig. 5.5. Stages are numbered from the feed end. Thus the final extract is E1 and the final raffinate is RN . The feed consists of carrier and solute, but can also contain solvent up to the solubility limit. Besides solute, the solvent can also contain carrier. Equilibrium is assumed to be achieved at each stage, so that for each stage the extract and the raffinate are in equilibrium for all three components. Because most liquid-liquid equilibrium data are given in mass concentrations we use the following symbol definitions: F S En Rn yi,n xi,n

= = = = = =

mass mass mass mass mass mass

flow rate of feed to the extractor flow rate of solvent to the extractor flow rate of extract leaving stage n flow rate of raffinate leaving stage n fraction of species i in extract leaving stage n fraction of species i in raffinate leaving stage n

The first step is to locate the mixing point M, which represents the overall composition of the combination of feed, F, and entering solvent, S in Fig. 5.7. The overall composition is obtained from the overall mass balance and the three material balances over each component: 3

See appendix to this chapter how to read compositions from ternary diagrams.

126

5 Liquid-liquid extraction

M=F+S

(5.34)

xi, M M = xi, F F + yi, S S

(5.35)

and

From any two of these xi,M values, point M is located as shown in Fig. 5.7. Based on the properties of the triangular diagram, point M must be located somewhere on the straight line connecting F and S. With point M in place, the exiting extract composition E1 is determined from overall mass balances: M = RN + E1

(5.36)

xi, M M = xi, Rn Rn + yi, E1 E1

(5.37)

and

Because the raffinate, RN , is assumed to be at equilibrium, its composition must lie on the equilibrium curve of Fig. 5.7. Therefore, if we specify the value xsolute,Rn , we can locate the point Rn . A straight line drawn from Rn through M will locate E1 at the intersection of the equilibrium curve, from which the composition of E1 can be read. Values of the flow rates Rn and E1 can then be determined from the overall material balances above or from Fig. 5.7 by the inverse leverarm rule: ¯ MRN ¯ E1 RN

(5.38)

¯ ME1 =¯ M E1 RN

(5.39)

E1 M

=

RN

The second step involves the determination of the operating point and operating lines. Analogous to the McCabe-Thiele method an operating line is the locus of passing streams. Referring to Fig. 5.5, material balances around groups of stages from the feed end are: F − E1 = ... = Rn − 1 − En = ... = RN − S = P

(5.40)

Because the passing streams are differenced, point P defines a difference point rather than a mixing point. From the geometric considerations as apply to a mixing point, a difference point also lies on a straight line through the points involved. However, while the mixing point always lies inside the triangular diagram and between the two end points, the difference point usually lies outside the triangular diagram along an extrapolation of the line through two points such as F and E1, RN and S and so on. To locate the difference point, two straight lines are drawn through the point pairs (E1, F) and (S, RN), which were established during the first step and also shown in Fig. 5.8. These lines are extrapolated until they intersect at

5.3 Design of countercurrent extractions

127

Fig. 5.8 Location of operating point and construction of operating lines.

point P. Fig. 5.8 shows these lines and the difference point P. From point P straight lines can be drawn through points on the triangular diagram for any other pair of passing streams. Thus, we refer to the difference point as an operating point, and the lines drawn through pairs of points for passing streams and extrapolated to point P as operating lines. The operating point P lies on the feed or raffinate side of the triangular diagram in the illustration of Fig. 5.8. Depending on the relative amounts of feed and solvent and the slope of the tie lines, point P may be located on the solvent or feed side of the diagram, and inside or outside the diagram. The third step involves the construction of additional tie lines that connect opposite sides of the equilibrium curve, which is divided into the two sides by the plait point. Typically a diagram will not contain all the tie lines needed. Tie lines may be added by centering them between existing experimental or predicted tie lines. With these additional tie lines the equilibrium stage construction can be started. Because the lines through point P represent the mass balances around each stage, Fig. 5.8 can now be used to determine the number of required equilibrium stages. Equilibrium stages are stepped off on the triangular diagram by alternate use of equilibrium and operating lines, as shown in Fig. 5.9. The approach is as follows. The extract stream E1 comes from the first stage and is in equilibrium with raffinate stream R1. This point must therefore be located at the opposite end of a tie line connecting to E1 representing the equilibrium relation in the ternary diagram. The extract stream of the second stage E2 passes R1 and is therefore found by the intersection of the operating line, drawn through points R1 and P. From E2 we continue by applying the equilibrium tie lines to locate R2 and determine E3 via the operating line. Continuing in this fashion by alternating between equilibrium tie lines and operating lines, we finally reach or pass the known point RN . If the latter, a fraction of the last stage is taken. The graphical procedure just described presupposes that a solvent-to-feed ratio was chosen that allows separation with a finite number of stages. As with distilla-

128

5 Liquid-liquid extraction

Fig. 5.9 Determination of the number of equilibrium stages.

Fig. 5.10

Determination of minimum solvent-to-feed ratio.

tion, this solvent-to-feed ratio cannot be chosen completely arbitrarily. This can be illustrated graphically by considering the previous system in the construction that is shown in Fig. 5.10. The points F, S and RN are again specified, but E1 is not because the solvent rate has not yet been specified. The first operating line is again drawn through the points S and RN and extended to the left and right of the diagram. This line is the locus of all possible material balances determined by adding S to RN . In order to minimize the amount of solvent, one needs to choose E1 as high as possible. For a given feed F, the highest possible value is found when the second operating line through point F coincides with a tie line. In Fig. 5.10 this is represented by the operating line drawn through point P = Pmin and point F extended to Emax at the intersection with the extract side of the equilibrium curve. Under this minimum solvent-to-feed ratio condition Emax should be in equilibrium with R1 but the operating line dictates that the extract is in equilibrium with E2. In this way we end up with a situation where the entering flow F and the exit flow Emax are in equilibrium. Hence an infinite amount of stages are required to achieve the desired separation.

5.3 Design of countercurrent extractions

129

From the compositions of the four points, S, RN , F and Emax, the mixing point M can be determined and the following material balances can then be used to calculate Smin /F: F + Smin = RN + Emax = M

xi, M M = xi, F F + yi, S Smin

and

(5.41)

For the extraction to be conducted in a finite number of stages, a solvent flow rate greater than Smin must be selected. A reasonable value for S might be 1.5 · Smin. An important second condition is that the mixing point M must lie in the two-phase region. As this point is moved along the line SF toward S, the ratio S/F increases according to the inverse lever-arm rule. In the limit, a maximum S/F ratio is reached when M = Mmax arrives at the equilibrium curve on the extract side. At this point all of the feed is dissolved in the solvent and no raffinate is obtained. To avoid this situation a solvent to feed ratio S/F must be selected such that: S

(F )

min


50 nm

εpore

ρpore g/cm3

S m2/g

2

3–6 25 2–4

0.5 0.5-0.7 0.4–0.6

1.2–1.3 0.6–1.1 0.5-0.9

200–400 200–900 400–1200

0.2–0.5

0.7 0.9–1.0

100–300

0.4–0.6

100–700

1 0.4–0.5

200–600 800 30–1000 10–100

25–40

0.4–0.6

applications for various gases. Like with silica gels, the water bond with the alumina surface is not as strong as with zeolites, so that regeneration of aluminas can often be accomplished at somewhat milder temperatures. As can be seen in Fig. 6.4, activated alumina can have a higher ultimate water capacity, but the zeolites have a higher capacity at low water partial pressures. Thus zeolites are typically chosen when very high water removal is necessary, while activated alumina is preferred if adsorbent capacity is more important.

Fig. 6.4 Comparison of water adsorption on various adsorbents.

6.2 Adsorption fundamentals

149

Fig. 6.5 Structure of styrene-divinylbenzene copolymer adsorbents.

Fig. 6.6 Schematic diagrams of structures of two common molecular sieve zeolites.

Polymer-based adsorbents are presently used in a few operations removing organic constituents from vent gas streams, such as the removal of acetone from air. As illustrated by Fig. 6.5, these materials are usually styrene-divinylbenzene copolymers, which in some cases have been derivatized to give the desired adsorption properties. In addition to equilibrium adsorption, selectivity of adsorbents may also originate from two important other separating mechanisms: • exclusion of certain molecules in the feed because they are too large to fit into the pores of the adsorbent (molecular sieving) • differences in the diffusion of different adsorbing species in the pores of the adsorbent (kinetic effect). Significant kinetic selecitivity is a characteristic feature of molecular sieve adsorbents such as carbon molecular sieves and zeolites. Molecular sieve zeolites are highly crystalline alumino-silicate structures with highly regular channels and cages as displayed in Fig. 6.6. Zeolites are selective for polar, hydrophilic species, and very strong bonds are created with water, carbon dioxide, and hydrogen sulfide while weaker bonds are formed with organic species. Molecular sieving is

150

6 Adsorption and ion exchange

Fig. 6.7 Comparison of molecular dimensions and zeolite pore sizes.

possible when a channel size is of molecular dimensions, restricting the diffusion sterically. As a result small differences in molecular size or shape can lead to very large differences in diffusivity. In the extreme certain molecules or a whole class of compounds may be completely excluded from the micropores. This is illustrated in Fig. 6.7, where a range of molecular sizes is compared with the channel diameters of various zeolites. The most important examples of such processes are the separation of linear hydrocarbons from their branched and cyclic isomers using a 5A zeolite adsorbent and air separation over carbon molecular sieve or 4A zeolite, in which oxygen, the faster diffusing component, is preferentially adsorbed. 6.2.2 Equilibria In adsorption a dynamic equilibrium is established for the distribution of the solute between the fluid and the solid surface, see Fig. 6.8. This equilibrium is commonly expressed in the form of an isotherm, which is a diagram showing the variation of the equilibrium adsorbed-phase concentration or loading with the fluid-phase concentration or partial pressure at a fixed temperature. For pure gases, experimental physical adsorption isotherms, in amount adsorbed q vs. partial pressure p or concentration c have shapes that are classified into five types by Brunauer, as shown in Fig. 6.9. Both Types I and II are desirable isotherms,

6.2 Adsorption fundamentals

151

Fig. 6.8 Schematic of the dynamic adsorption/desorption equilibrium (fraction occupied θ = 6 / 14; fraction not occupied 1 – θ = 8 / 14).

Fig. 6.9 The Brunauer classification of adsorption isotherms.

exhibiting strong adsorption, even at lower concentrations. The simplest and still the most useful isotherm is the Langmuir Type I isotherm: θA =

qA qmax, A

=

bA pA 1 + bA pA

=

bA′ cA 1 + bA′ cA

where θ = fraction of sites occupied (see Fig. 6.8), q = amount adsorbed per unit volume or unit weight adsorbent,

(6.4)

152

6 Adsorption and ion exchange

qmax = monolayer capacity, referring to maximum occupancy or θ = 1, and b = Langmuir adsorption constant in reciprocal pressure or concentration units. Shown in Fig. 6.9 are the two limiting cases of a type I isotherm: linear behavior at low (relative) pressures (bA pA ≪ 1) and maximum occupancy at higher concentrations (bA pA ≫ 1). For bA pA ≪ 1, the Langmuir isotherm reduces to a linear form and the following form of Henryʼs law1, called a linear isotherm, is obeyed: qA = qmax, A bA pA

(6.5)

The derivation of the Langmuir expression is based on four assumptions: 1. 2. 3. 4.

The adsorbed molecules or atoms are held at definite, localized sites Each site can accommodate only one molecule or atom The energy of adsorption is constant over all sites, and There is no interaction between neighboring adsorbates

Though the detailed derivation of Eq. 6.4 follows from sound thermodynamic reasoning, it can also be obtained by realizing that adsorption is the result of fast and simultaneous adsorption and desorption and that at equilibrium the rates of adsorption and desorption are equal. The rate of adsorption, rads , is proportional to the partial pressure, pA as well as the fraction of empty sites, 1 − θ, whereas the rate of desorption, rdes , is proportional to the fraction of occupied sites, θ. At equilibrium: rads − rdes = kads pA (1 − θ) − kdes θ = 0

(6.6)

where kads is a first order rate constant and kdes a zero order rare constant. Rearrangement of Eq. 6.6 and introducing bA =

kads

(6.7)

kdes

gives the Langmuir equation for adsorption from the gas phase in non-dimensional form as in Eq. 6.4; it often takes the following dimensional form: qA = qmax, A

bA pA 1 + bA pA

(6.4b)

The adsorption constant bA is temperature dependent in a way that is related to the activation energy of adsorption through an Arrhenius type equation: bA = bA0 e −ΔEads ∕R T

(6.8)

For an exothermic process ΔEads is negative, and the adsorption constant decreases with increasing temperature. 1

Note that qmax · bA equals the Henry coefficient HA in Eq. 1.9a.

6.2 Adsorption fundamentals

153

The Langmuir model is the correct form to represent a type I isotherm, since at low pressure it approaches Henryʼs law while at high pressure it tends asymptotically to the saturation limit. Although only few systems conform exactly to the Langmuir model, it provides a simple qualitative representation of the behavior of many systems and it is therefore widely used. For adsorption from a liquid phase, it is convenient to replace the (partial) pressure by concentration: qA = qmax,A

bA′ cA 1 + bA′ cA

(6.9)

where b ′ the Langmuir adsorption constant in reciprocal concentration units. The assumptions mentioned above for adsorption from the gas phase are hardly valid for adsorption from (aqueous) solutions and the physical meaning of qmax,A and bA′ is lost. In many cases, however, the equation is still applicable; both parameters can be treated as empirical constants, qmax and bA′, obtained by fitting the nonlinear equation directly to experimental data or by employing a linearized form of Eq. 6.4b: cA qA

=

1 qmax bA′

+

1 qmax

cA

(6.10)

Using Eq 6.10, the best straight line is drawn through a plot of cA /qA versus cA, giving a slope of 1/qmax and an intercept of 1/qmax bA′. The Langmuir model can be simply extended to multicomponent systems, reflecting the competition between species for the adsorption sites: qA qmax, A

=

bA pA 1 + bA pA + bB pB + .......

(6.11)

and qB qmax, B

=

bB pB 1 + bA pA + bB pB + .......

(6.12)

Since the total number of available adsorption sites should be equal for each component, thermodynamic consistency requires qmax,A = qmax, B. In practice this requirement can cause difficulties when correlating data for sorbates of very different molecular size. For such systems it is common practice to ignore this requirement, thereby introducing qmax, B as an additional model parameter. The ′ (see Eq. 1.10), now simply correequilibrium selectivity of the adsorbent, αAB sponds to the ratio of the equilibrium constants from the multicomponent Langmuir model:

′ = αAB

qA pA qB pB

=

bA bB

(6.13)

154

6 Adsorption and ion exchange

Because this selecitivity is independent of composition the ideal Langmuir model is often referred to as the constant separation factor model. 6.2.3 Kinetics In general the rate of physical adsorption is controlled by diffusional limitations rather than by the actual rate of equilibration inside the adsorbent particle at the pore surface, which is a very fast process. As illustrated schematically in Fig. 6.10, the adsorption of a solute onto the porous surface of an adsorbent requires the following steps: 1. External mass transfer of the solute from the bulk fluid through a thin film or boundary layer, to the outer solid surface of the adsorbent. 2. Internal mass transfer of the solute by pore diffusion from the outer surface of the adsorbent to the inner surface of the internal pore structure. 3. Surface diffusion along the porous surface. 4. Adsorption of the solute onto the porous surface. During regeneration of the adsorbent, the reverse of the four steps occurs. However, under most practical conditions the external film resistance is seldom rate limiting, so that internal mass transfer generally controls the adsorption/desorption rate. From this perspective adsorbents may be divided into two broad classes: homogeneous and composite. Fig. 6.11 illustrates that in homogeneous adsorbents (activated alumina, silica gel) the pore structure persists throughout the particle, while the composite adsorbent particles (zeolites, carbon molecular sieves) are

Fig. 6.10 [42]).

Resistances to mass transfer in a composite adsorbent pellet (adapted from

6.2 Adsorption fundamentals

155

Fig. 6.11 Homogeneous and composite microporous adsorbent particle (adapted from [42]).

formed by aggregation of small microporous microparticles. As a result the pore size distribution in composite particles has a well-defined bimodal character with micropores within the microparticles connected through macropores within the pellet. Transport in a macropore can occur by bulk molecular diffusion, Knudsen diffusion, surface diffusion and/or Poiseuille flow. In liquid systems bulk molecular diffusion is generally dominant, but in the vapor phase the contributions from Knudsen and surface diffusion may be large or even dominant. Knudsen diffusion becomes dominant when collisions with the pore wall occur more frequently than collisions with diffusing molecules. In micropores the diffusing molecule never escapes from the force field of the pore wall and the Knudsen mechanism no longer applies. Diffusion occurs by jumps from site to site, just as in surface diffusion, and the diffusivity becomes strongly dependent on both temperature and concentration. The selectivity in a kinetically controlled adsorption process depends on both kinetic and equilibrium effects. This kinetic selectivity can be approximated when two species (A and B) diffuse independently with diffusivites DA and DB, respectivily and their isotherms are also independent. Under these conditions the ratio of their uptakes at any time will be given by: qA pA αAB =

qB pB

=

bA DA bB √ DB

(6.14)

A typical example is shown in Fig. 6.12, where the equilibrium adsorption and loading curves of oxygen and nitrogen on carbon molecular sieves is shown. The dramatic difference between the uptake rates is caused by the fact that the pore diameter of the carbon is just very slightly larger than the diameters of the two

156

6 Adsorption and ion exchange

Fig. 6.12 Equilibrium adsorption and uptake rates for Carbon Molecular Sieves used in air separation.

gases. As a result nitrogen has much more difficulty traversing a pore than oxygen. Thus, even though there is virtually no equilibrium selectivity, the operation in the kinetic region results in very high selectivity for oxygen/nitrogen separation.

6.3 Fixed-bed adsorption 6.3.1 Bed profiles and breakthrough curves In a fixed bed a nearly solute free liquid or gas effluent can be obtained until the adsorbent in the bed approaches saturation. Under the conditions that external and internal mass-transfer resistances are very small, plug flow is achieved, axial dispersion is negligible, the adsorbent is initially free of adsorbate and the adsorption isotherm begins at the origin, equilibrium between the fluid and the adsorbent is achieved instantaneously. Under these ideal fixed bed adsorption conditions the instantaneous equilibrium results in a shock like wave, the stoichiometric front (Fig. 6.13), moving as a sharp concentration step through the bed. Upstream of the front, the adsorbent is saturated with adsorbate and the concentration of solute in the fluid is that of the feed. In the upstream region the adsorbent is spent. Downstream of the stoichiometric front and in the exit fluid, the concentration of the solute in the fluid is zero, and the adsorbent is still adsorbate free. After a period of time, called the stoichiometric time, the stoichiometric wave front reaches the end of the bed and the concentration of the solute in the fluid rises abruptly to the inlet value. Because no further adsorption is possible this point is referred to as the break through point. In a real fixed-bed adsorber internal and external transport resistance are finite and axial dispersion can be significant. These factors contribute to the development of broader S-shaped concentration fronts like that shown in Fig. 6.14. At

6.3 Fixed-bed adsorption

157

Fig. 6.13

Stoichiometric equilibrium concentration front for ideal fixed-bed adsorption.

Fig. 6.14

Breakthrough curve for real fixed-bed adsorption.

the inlet of the bed the adsorbent has become saturated and in equilibrium with the adsorbate in the inlet fluid. At the exit end, the adsorbate content is still at its initial value. In between there is a reasonably well-defined mass transfer zone in which the adsorbate concentration drops from the inlet to the exit value. This zone progresses through the bed as the run proceeds. At the breakpoint time tbp the zone begins to leave the column. If the run is extended the effluent concentra-

158

6 Adsorption and ion exchange

tion rises sharply until the whole bed is saturated, tsat , and the effluent concentration equals the inlet concentration. The S-shaped curve is called the breakthrough curve, representing the outlet-to-inlet solute concentration in the fluid as a function of time from the start of the flow. The region between tbp and tsat is the Mass-Transfer-Zone, MTZ, where adsorption takes place. At the breakthrough point tbp , the leading point of the MTZ just reaches the end of the bed. For a flow rate Q, the useful capacity of the bed, Quc , and its total capacity Qtot are respectively: tbp

∫ Q (c − c

dt

(6.15)

Qtot = Q (cf − cout) dt

(6.16)

Q uc =

f

out)

0

and

∞

∫ 0

6.3.2 Equilibrium theory model The steepness of the breakthrough curve determines the extent to which the capacity of an adsorbent bed can be utilized. Another complication is the fact that the steepness of the concentration profiles increases or decreases with time, depending on the shape of the adsorption isotherm. Therefore, determination of the shape of the curve is very important in determining the length of an adsorption bed. For the fixed bed shown in Fig. 6.15 the velocity of the concentration front can be derived by considering ideal plug flow of fluid through the column cross sectional area, A, at a constant volume flow rate Q and constant temperature. As a first approximation the so-called equilibrium model is used, which assumes that, at any point in the bed, equilibrium between the fluid phase and the adsorbent phase is achieved instantaneously. Under these conditions the mass balance in a volume element, A Δz, of the bed with interparticle porosity εbed over a time step Δt becomes: accumulation in solution + accumulation in porous solid = flow in − flow out or AΔz εbed

Δc Δt

+ AΔz (1 − εbed) ρpart

Δq Δt

= Qc − Q (c + Δc)

(6.17)

where ρpart = particle density (Fig. 6.3) and q = amount adsorbed per unit weight of adsorbent particles. Dividing by AΔz and taking the limit as Δz → 0 gives εbed

dc dt

+ (1 − εbed) ρpart

dq dt

=−

Q dc A dz

(6.18)

6.3 Fixed-bed adsorption

159

Concentration change in a thin slice of a fixed bed adsorber.

Fig. 6.15

By the introducing the superficial velocity ucarrier ≡ Q/A, the chain rule and the rules of implicit partial differentiation:

∂q ∂t

=

∂q ∂c ∂c ∂t

and

∂z ∂t

|

∂c ∂t =− c

∂c ∂z

| |

z

(6.19)

t

the velocity vc of propagation of a concentration c becomes: vc ≡

∂z ∂t

|

u carrier

= c

ε bed

dq + (1 − ε bed) ρpart dc

|

≈ c

u carrier dq (1 − ε bed) ρpart dc

|

(6.20) c

because usually εbed ≪ (1 − εbed) · ρpart · dq/dc. This equation gives the velocity vc of any solute concentration c in the concentration wave front in terms of the superficial fluid velocity ucarrier , and the slope of the adsorption isotherm dq/dc at that concentration. In the analysis of adsorption and desorption processes Eq. 6.20 is important because it illustrates for the equilibrium case in which way the shape of the adsorption wave changes as it moves along the bed. For a linear isotherm dq/dc is constant and all concentrations move at the same velocity vc. If the adsorption isotherm is curved, regions of the front at a higher concentration move at a velocity different from regions at a lower concentration. For these three conditions, Fig. 6.16 illustrates the development of displacement of concentration gradients. For unfavorable isotherms (Fig. 6.16a) the concentration gradients in the column at various times show that the wave front is dispersive, and leads to an adsorption zone which gradually increases in length as it moves through the bed. For the case of a linear isotherm the zone goes through the bed unchanged. For favorable isotherms such as the Langmuir type (Fig. 6.16c), concentration profiles become sharper and the wave front is compressive because points of high concentration in the adsorption wave move more rapidly than points of low concentrations. Since it is physically impossible for

160

6 Adsorption and ion exchange

Fig. 6.16 Development of an adsorption wave through a bed for (a) unfavorable; (b) linear and (c) favorable isotherms. The mass transfer zone at a certain time and increasing (a); constant (b) and decreasing (c) in time.

points of high concentration to overtake points of low concentration, the effect is for the adsorption zone to become narrower as it moves along the bed. The physical limit is a shock wave with a shock wave velocity: vsw ≈

ucarrier

(6.21)

Δq (1 − εbed) ρpart Δc

where Δq and Δc are calculated between the feed and pre-saturation conditions. The time tsat required saturating a freshly regenerated adsorber of length L with a feed concentration cf is the time of movement of this shock wave through the column: tsat =

L vsw

≈

L ucarrier

⋅ (1 − εbed) ρpart

q∗ − 0 cf − 0

(6.22)

where q* = amount adsorbed in equilibrium with feed concentration cf . Note that Eq. 6.22 can also be derived from a mass balance over the entire column (adsorbent weight Ws , plug flow Q) when the column is just saturated with c = cf at t = tsat : tsat Q cf = Ws q ∗

(6.23)

6.3 Fixed-bed adsorption

161

6.3.3 Modeling of mass transfer effects In a real system the finite resistance to mass transfer and axial mixing in the column lead to departures from the idealized response predicted by the equilibrium theory. In the case of a favorable isotherm the concentration profile spreads in the initial region until a stable situation is reached in which the mass transfer rate is the same at all points along the wave front and exactly matches the shock velocity. This represents a stable situation and the profile propagates without further change in shape, as illustrated in Fig. 6.17. The distance required to approach the constant pattern limit decreases as the mass transfer resistance decreases and the non-linearity of the equilibrium isotherm increases. Establishing the form of the concentration profile under these constant pattern conditions requires extension of the model with mass transfer resistances outside and inside the adsorbent particle. A model that includes all these mass transfer steps requires simultaneous solutions of the mass balance equations within the particle as well as in the bed or for the bulk flow. This simultaneous solution is a formidable task, which can be avoided by using the linear driving force (LDF) model. In this model, which is widely used to simulate and design fixed-bed adsorption, the mass balance equation within the particle is eliminated by relating the overall uptake rate dqA /dt in the particle to the bulk flow concentration cA. This way the particle mass balance equation is eliminated from the model and the rate of mass transfer may be expressed in terms of a simple linear rate expression: ∂q¯A ∂t

= f (qA∗) = k (qA∗ − q¯A)

(6.24a)

where q¯A = the average or overall amount adsorbed in the particle qA* = the amount adsorbed in equilibrium with the adsorbate concentration cA. k = 1st order rate constant The rate constant k includes both external and internal transport resistances: 1 k

=

a kf

+

2 dpart

60 Deff

(see, e.g., Henley et.al. [10] for details)

(6.24b)

Fig. 6.17 Effect of mass transfer on the adsorption wave through a bed for a favorable isotherm.

162

6 Adsorption and ion exchange

where kf = external mass transfer coefficient, a = Vpart /Apart = dpart /6 (spherical particles) Deff = effective coefficient of diffusion inside particle Incorporation of Eqs. 6.24 into the previously derived mass balance (Eq. 6.18) and taking into account axial dispersion (represented by the term with the coefficient of axial dispersion, Dax) gives: k (1 −

εbed) ρpart (qA∗

− q¯A) = − ucarrier

∂cA ∂z

+ Dax

d 2 cA dz 2

(6.25)

This equation provides the basic relation that can be integrated to calculate the effect of mass-transfer resistances on the shape of breakthrough curves according to the LDF-model. The self-sharpening effect as depicted in Fig. 6.16c is balanced by axial dispersion; the resulting constant pattern is shown in Fig. 6.17.

6.4 Basic adsorption cycles Commercial adsorptions can be divided into bulk separations, in which about 10 wt% or more of a stream must be adsorbed, and purifications, in which usually considerably less than 10 wt% of a stream must be adsorbed. Such a differentiation is desirable because in general different process cycles are used for the two categories. As schematically illustrated in Fig. 6.18, there are three basic ways to influence the loading of an adsorbate on an adsorbent: 1. Changing temperature 2. Changing (partial) pressure in a gas or concentration in a liquid 3. Adding a component which competitively adsorbs with the adsorbate of interest A major difference in processes based on temperature change and those based on pressure or concentration change is the time required to change the bed from an adsorbing to a desorbing or regenerating condition. In short, pressure and concentration can be changed much more rapidly than temperature. The consequence is that temperature-swing processes are limited almost exclusively to low adsorbate concentrations in the feed. For higher adsorbate feed concentrations the adsorption time would become to short compared to the regeneration time. In contrast, pressure-swing or concentration-swing processes are much more suitable for bulk separations because they can accept feed, change to regeneration, and be

6.4 Basic adsorption cycles

Fig. 6.18

163

Schematic illustration of adsorption/desorption cycles.

Fig. 6.19 Dispersive wave-front and breakthrough curve for desorption caused by a favorable isotherm.

back to the feed condition within a reasonable fraction of the cycle time. The primary quantities of interest in regeneration are the required desorption time, tdes , to achieve a desired outlet concentration and the corresponding purge feed consumption. Desorption of a species according to a Langmuir isotherm is equivalent to the adsorption a species with unfavorable isotherm. The input is now a negative step, which is shown in Fig. 6.19 together with the concentration profiles in the bed and the desorption breakthrough curve. Because the high concentrations travel faster than the low concentrations, the breakthrough curve is now of a dispersive nature. The desorption time, tdes , is given by the bed height, L, divided by the propagation velocity of the wave front of the concentration at which the desorption takes place. According to the equilibrium theory, it follows from Eq. 6.20:

164

6 Adsorption and ion exchange

tdes =

L vcdes

≈

L ucarrier

For desorption with cdes = 0,

(1

dq dc

− εbed) ρpart

|

dq dc

|

(6.26) cdes

represents the initial slope of the isotherm. c=0

6.4.1 Temperature-swing A temperature-swing or thermal-swing adsorption (TSA) cycle is one in which desorption takes place at a temperature much higher than adsorption. In this cycle, shown in Fig. 6.20, a stream containing a small amount of an adsorbate at concentration c1 is passed through the adsorbent bed at temperature T1 . After equilibrium between adsorbate in the feed and on the adsorbent is reached, the bed temperature is raised to T2 , and desorption occurs by a continued feed flow through the bed. In general the regeneration is more efficient at higher temperatures. Heating, desorbing and cooling of the bed usually takes in the range of a few hours to over a day, making the desorption step often the cost-determining factor in the separation. Therefore, temperature-swing processes are used almost exclusively to re-

Fig. 6.20 Temperature-swing adsorption (TSA).

6.4 Basic adsorption cycles

165

move small concentrations of adsorbates in order to maintain the on-stream time a significant fraction of the total process cycle time. The most common application of TSA is drying. Zeolites, silica gel and activated aluminas are widely used in the natural gas, chemical and cryogenics industries to dry streams. Of these adsorbents silica gel requires the lowest temperatures for regeneration. Other TSA applications range from CO2 removal to hydrocarbon separations, and include the removal of pollutants, odors and contaminants with activated carbon. 6.4.2 Pressure-swing In a pressure-swing adsorption (PSA) cycle desorption is effected at a pressure much lower than adsorption. The most common processing scheme has two, illustrated in Fig. 6.21, or three fixed bed adsorbers alternating between the adsorption step and the desorption steps. A complete cycle consists of at least three steps: adsorption, blow down and repressurization. During adsorption, the less selectively adsorbed components are recovered as product. The selectively adsorbed components are removed from the adsorbent by adequate reduction in partial pressure. During this countercurrent depressurization (blow down) step, the strongly adsorbed species are desorbed and recovered at the adsorption inlet of the bed. The repressurization step returns the adsorber to feed pressure and can be carried out with product or feed. PSA cycles operate at nearly constant temper-

Fig. 6.21

Pressure-swing adsorption (PSA).

166

6 Adsorption and ion exchange

ature and require no heating or cooling steps. They utilize the exothermic heat of adsorption remaining in the adsorbent to supply the endothermic heat of desorption. The principle application of PSA is for bulk separations where contaminants are present at high concentration. Fortunately, packed beds of adsorbent respond rapidly to changes in pressure. In most applications equilibrium adsorption is used to obtain the desired separation. However, in air separation PSA processes the adsorptive separation is based on kinetically limited systems where oxygen is preferentially adsorbed on 4A zeolite when the equilibrium selectivity favors N2 adsorption. The major purification applications for PSA are for hydrogen, methane and drying. Air separation, methane enrichment and iso-/normal separations are the principal bulk separations for PSA. 6.4.3 Inert and displacement purge cycles In a purge swing adsorption cycle the adsorbate is removed by passing a nonadsorbing gas or liquid, containing very little to no adsorbate, through the bed. Desorption occurs because the partial pressure or concentration of the adsorbate around the particles is lowered. If enough purge gas or liquid is passed through the bed, the adsorbate will be completely removed. Most purge-swing applications use two fixed-bed adsorbers to provide a continuous flow of feed and product (Fig 6.22). Its major application is for bulk separations when contaminants are at

Fig. 6.22 Inert purge cycle.

6.5 Principles of ion exchange

Fig.

6.23

167

Displacement purge cycle.

high concentration. Applications include the separation of normal from branched and cyclic hydrocarbons, gasoline vapor recovery and bulk drying of organics. The displacement purge-cycle differs from the inert-purge cycle in that a gas or liquid that adsorbs about as strongly as the adsorbate is used to remove the adsorbate. Desorption is thus facilitated by adsorbate partial pressure or concentration reduction in the fluid around the particles in combination with competitive adsorption of the displacement medium. The use of an adsorbing displacement purge fluid causes a major difference in the process. Since it is actually absorbed, it will be present when the adsorption cycle begins and therefore contaminate the less-adsorbed product. This means that the displacement purge fluid must be recovered from both product streams, as illustrated by Fig. 6.23.

6.5 Principles of ion exchange 6.5.1 Ion exchange resins Naturally occurring inorganic alumino-silicates (zeolites) were the first ion exchangers used in water softening. Today, synthetic organic polymer resins based on styrene or acrylic acid type monomers are the most widely used ion exchangers. These resin particles consist of a three-dimensional polymeric network with

168

6 Adsorption and ion exchange

Fig. 6.24 Styrene and divinyl benzene based ion exchange resins: (a) sulfonated cation exchanger; (b) aminated anion exchanger.

attached ionic functional groups to the polymer backbone. Ion exchange resins are categorized by the nature of functional groups attached to a polymeric matrix, by the chemistry of the particular polymer in the matrix. Strong acid and strong base resins are based on the copolymerization of styrene and a cross-linking agent, divinylbenzene, to produce the three dimensional cross-linked structure as shown in Fig. 6.24. In strong acid cation exchange resins sulfonic acid groups in the hydrogen form exchange hydrogen ions for the other cations present in the liquid phase: resin-SO3– H+ + Na+ ⇔ resin-SO3– Na+ + H+

(6.27)

It is not always necessary for the resin to be in the hydrogen form for adsorption of cations. Softening of water is accomplished by displacing sodium ions from the resin by calcium ions, for which the resin has a greater affinity: 2 resin-SO3– Na+ + Ca2+ ⇔ (resin-SO3– )2Ca2+ + 2 Na+

(6.28)

In strong base anion exchange resins quaternary ammonium groups are used as the functional exchange sites. They are used most often in the hydroxide form to reduce acidity: resin-N+(CH3)3 OH– + H+ + Cl– ⇔ resin-N+(CH3)3 Cl– + H2O (6.29) During the exchange the resin releases hydroxide ions as anions are adsorbed from the liquid phase. The effect is elimination of acidity in the liquid and conversion of the resin to a salt form. Ion exchange reactions are reversible. After complete loading a regeneration procedure is used to restore the resin to its original ionic form. For strong cation and anion resins this is typically done with dilute (up to 4 %) solutions of hydrochloric acid, sulfuric acid, sodium hydroxide or sodium chloride. Weak acid cation exchanger resins have carboxylic acid groups attached to the polymeric matrix derived from the copolymerization of acrylic acid and metha-

6.5 Principles of ion exchange

169

crylic acid. Weak base anion exchanger resins may have primary, secondary or tertiary amines as functional groups. The tertiary amine is most common. Weak base resins are frequently preferred over strong base resins for removal of strong acids in order to take advantage of the greater ease of regeneration. 6.5.2 Equilibria and selectivity Ion exchange differs from adsorption in that one sorbate (a counterion) is exchanged for a solute ion, and the exchange governed by a reversible, stoichiometric chemical reaction equation. The exchange equilibria depend largely on the type of functional group and the degree of cross-linking in the resin. The quantity of ions, acids or bases that can be adsorbed or exchanged by the resin is called the operating capacity. Operating capacities vary from one installation to another as a result of differences in composition of the stream to be treated. In ion exchange significant exchange does not occur unless the functional group of the resin has a greater selectivity for the ions in solution than for ions occupying the functional groups, or unless there is excess in concentration as in regeneration. This is pictured in the following reaction where the cation exchange resin removes cations B+ from solution in exchange for A+ on the resin: resin-SO3–A+ + B+ ⇔ resin-SO3– B+ + A+

(6.30)

For this case we can define a conventional chemical equilibrium constant, determining the selectivity for B over A:

Fig. 6.25

Equilibrium plot for univalent-univalent ion exchange.

170

6 Adsorption and ion exchange

KAB =

qB, resin mA, liquid qA, resin mB, liquid

(6.31)

where m and q are the concentrations of the ions in the solution and the resin phase, respectively. A typical plot of KAB for univalent exchange is given in Fig. 6.25. B is preferred by the exchanger if KAB > 1 while is KAB < 1 B is less preferred and the ion exchanger preferentially absorbs species A. Selectivity for ions of the same charge usually increases with atomic weight (Li < Na < K < Rb < Cs) and selectivities for divalent ions are greater than for monovalents. For practical applications it is rarely necessary to know the selectivity precisely. However, knowledge of relative differences is important when deciding if the reaction is favorable or not. Selectivity differences are marginally influenced by the degree of cross-linking of a resin. The main factor is the structure of the functional groups.

6.6 Ion exchange processes Ion exchange cycles consist of two principal steps, adsorption and regeneration. During adsorption impurities are removed or valuable constituents are recovered. Regeneration is in general of much shorter duration than the adsorption step. Industrial systems may be batch, semi continuous or continuous. They vary from simple one-column units to numerous arrays of cation and anion exchangers, depending upon the required application and quality of the effluent. A singlecolumn installation is satisfactory if the unit can be shut down for regeneration. However, for the processing of continuous streams, two or more columns of the same resin must be installed in parallel. Most ion exchanger columns are operated concurrently with the process stream and regeneration solution flowing down through the resin bed to avoid possible fluidization of the resin particles. Deionization processes require two columns, one containing a cation exchanger and one an anion exchanger. The cation exchanger unit must be a strong acidtype resin and it must precede the anion exchange unit. Placing the anion exchanger first generally causes problems with precipitates of metal hydroxides. A column containing a mixture of cation and anion exchanger resins is called a mixed bed. The majority of the used resins consist of strong acid cation and strong base anion exchangers. Mixed bed systems yield higher quality deionized water or process streams than when the same resins are used in separate columns. However, regeneration of a mixed bed is more complicated than of a two-bed system. To increase resin utilization and achieve high ion exchange reaction efficiency, numerous efforts have been made to develop continuous ion exchange contactors. Two examples of them are shown in Figs. 6.26 and 6.27. The Higgins Loop contactor operates as a moving packed bed by using intermittent hydraulic pulses

6.6 Ion exchange processes

Fig. 6.26

Higgins moving bed contactor (adapted from [15]).

Fig. 6.27

Himsley fluidized bed contactor (adapted from [10, 15]).

171

172

6 Adsorption and ion exchange

to move incremental portions of the bed from the contacting section to the regeneration section. In other continuous systems, such as the Himsley contactor, columns with perforated plates are used. The liquid is pumped into the bottom and flows upward through the column to fluidize the resin beads on each plate. Periodically, the flow is reversed to move incremental amounts of resin to the stage below. The resin at the bottom is lifted via the wash column to the regeneration column. A more recent approach involves the placement of a number of columns on a carousel that rotates constantly at an adjustable speed.

Nomenclature a A, Apart b c d Dax Deff k kf K L q

particle volume / surface area ratio (= dpart / 6 for spheres), Eq. 6.24b area (column cross section, particle interface area) adsorption constant, Eq. 6.4 concentration diameter coefficient of axial dispersion, Eq. 6.25 effective diffusivity in the pores of a porous solid, Eq. 6.24b 1st order rate constant, Eq. 6.24a external mass transfer coefficient, Eq. 6.24b equilibrium constant column length amount adsorbed

Q S t u v x, y V Ws z α ε ρ θ

volume flow rate surface area in porous solid, Eq. 6.2 time superficial velocity, Eq. 6.18 interstitial velocity, Eq. 6.20 mole fraction volume, Fig. 6.3 weigth of dry solid (adsorbent) length coordinate selectivity, Eq. 6.14 porosity, fractional void volume density occupancy, Eq. 6.4

Indices A, B ads bp c des f

components adsorption break point at concentration c desorption feed

[m3 m–2] [m2] [m3 mol, Pa–1] [mol m–3] [m] [m2 s–1] [m2 s–1] [s–1] [m s–1] [–] [m] [mol kg–1 (mol m–3)] [m3 s–1] [m2] [s] [m s–1] [m s–1] [–] [m3] [kg] [m] [–] [–] [kg m–3] [–]

Exercises

max part sat

173

maximum (monolayer) capacity particle at saturation condition

Exercises 1. The following table gives the Langmuir isotherm constants for the adsorption of propane and propylene on various adsorbents. Adsorbent

Adsorbate

qm (mmol/g)

b (1/bar)

ZMS 4A

C3 C3= C3 C3= C3 C3= C3 C3=

0.226 2.092 1.919 2.436 2.130 2.680 4.239 4.889

9.770 95.096 100.223 147.260 55.412 100.000 58.458 34.915

ZMS 5A ZMS 13X Activated Carbon

a. b. c. d.

Which component is most strongly adsorbed by each of the adsorbents Which adsorbent has the greatest adsorption capacity Which adsorbent has the greatest selectivity Based on equilibrium conditions, which adsorbent is the best for the separation

2. The following data were obtained for the adsorption of toluene from water on activated carbon and water from toluene on activated alumina. Fit both sets to a Langmuir-type isotherm and compare the resulting isotherms with the experimental data. Toluene (in water) on Activated Carbon

Water (in toluene) on Activated Alumina

c (mg/l)

q (mg/g)

c (ppm)

q (g/100g)

0.01 0.02 0.05 0.1 0.2 0.5 1 2 5 10

12.5 17.1 23.5 30.3 39.2 54.5 70.2 90.1 125.5 165.0

25 50 75 100 150 200 250 300 350 400

1.9 3.1 4.2 5.1 6.5 8.2 9.5 10.9 12.1 13.3

3. An aqueous amount of V m3 contains 0.01 kg/m3 of nitrobenzene. We want to reduce this to 10–6 kg/m3 by adsorption on activated carbon. The solution is fed to an ideally stirred tank containing mC kg adsorbent, see Fig. 6.28. At these low concentrations, the equilibrium adsorption isotherm appears to be linear:

174

6 Adsorption and ion exchange

q = H′ · c with q = amount adsorbed, kg/kg c = concentration of nitrobenzene, kg/m3 H ′ = Henry constant = 675 m3/kg How many kg of activated carbon is required per m3 of treated water? Fig. 6.28

4. Nitrobenzene in an aqueous effluent should be reduced from its initial concentration of 0.02 kg/m3 to a value of not more than 2 · 10–5 kg/m3. Nitrobenzene can be removed from water by adsorption on active carbon. The amount q of nitrobenzene adsorbed in equilibrium with its concentration c in water is given by the following Langmuir isotherm: q (c) =

510 ⋅ c 1 + 4550 ⋅ c

with c in kg/m3 and q in kg/kg

Two different process schemes (see outlines below) are considered to carry out the required reduction: – a packed column (see process scheme Fig. 6.29a below), and – a batch process in an ideally stirred tank (see process scheme Fig. 6.29b below)

Fig. 6.29

The packed column contains Wpc kg carbon and is fed with waste water at a rate of Φf = 0.015 m3 s–1. The rate of convective transport through the bed is such that local adsorption equilibrium is established throughout the column. The concentration profile is flat (plug flow), as long as the column is not saturated, the outlet concentration is zero. The length of the column L = 2.0 m, its diameter D = 0.6 m, the bed porosity εbed = 0.6 and the particle density ρp = 800 kg m–3. a1. Calculate from a component balance the amount of carbon per unit volume of waste water (Wpc / V in kg/m3)

Exercises

175

a2. How much waste water (V in m3) is purified when the column is about to break through? After saturation of the column, it is regenerated with a water flow of the same size in which the concentration of nitrobenzene amounts to 2 · 10–5 kg m–3. a3. How long does it take to regenerate the column? In the second process, an ideally stirred tank, V m3 waste water (as calculated in part a1) is treated. This batch is treated with Wbatch kg active carbon. After equilibration the resulting concentration of nitrobenzene should be 2 · 10–5 kg/m3. b1. Calculate the excess amount of carbon to be used in the batch process compared to that required in the continuous process, in other words, calculate Wbatch / Wpc . b2. Comment on the answer of part b1. 5. Derive the expression for the time required to saturate a fixed bed with c = cf (Eq. 6.22) starting from the overall mass balance given in Eq. 6.23. List all the necessary assumptions. 6. A couple of students got an assignment to determine experimentally the residence time of CO2 during adsorption and desorption in a packed bed column. They had to compare the results with a theoretical model based on local adsorption equilibrium. A column (L = 1 m, internal diameter D = 5 mm) was filled with activated carbon and its exit provided with a CO2-detector. The temperature was kept constant at 20 °C. Flushing with pure Helium pretreated the column. At time τ0 a CO2-He mixture, containing 4 vol% CO2 , was fed to the column. Some time after saturation, at time τ1, the feed was switched to pure He. The concentration CO2 in the effluent was measured continuously and is given schematically (not to scale) in the plot below (Fig 6.30):

Fig. 6.30

Additional data The particle density of the carbon ρparticle = 790 kg/m3 and the bed porosity εbed = 0.40. The adsorption of CO2 at 20 °C is given by a Langmuir-type adsorption isotherm: q = qm

bc CO 2 1 + bc CO 2

where monolayer capacity qm = 1.39 mol/kg, the adsorption constant b = 0.068 m3/ mol and the concentration c is in mol m–3 In all experiments (flushing with He, adsorption with He/CO2 , desorption with He) the total gas flow ΦV was maintained at 4 · 10–6 m3/s at 20 °C and 1 bar. In this equipment and at these conditions, adsorption and desorption are isothermal processes. R = 8.31 J/mol K.

176

6 Adsorption and ion exchange

The following problems are to be addressed: a. Explain the form of the breakthrough curve qualitatively. b. Calculate the breakthrough time, e.g. τbreakthrough − τ0, using Eq. 6.22. c. Some students calculated the breakthrough time from the ratio of the maximum amount to be adsorbed and the flow rate of carbon dioxide: τbreakthrough − τ0 =

Ws ⋅ q (cf) Φv ⋅ cf

When calculated properly, this gives the same result as in part b. Explain. d. Calculate the desorption time, τ∞ − τ1. 7. A commercial ion-exchange resin is made of 88 wt% styrene (MW = 0.104 kg/mol) and 12 wt% divinyl benzene (MW = 0.1302 kg/mol). Estimate the maximum ionexchange capacity in equivalents/kg resin when an sulfonic acid group (MW = 0.0811 kg/mol) has been attached to each benzene ring. 8. A continuous stream of a soil/water mixture (0.3 m3/s) containing 15 mol/m3 of a heavy metal M+ is treated with an ion exchange resin to remove the heavy metal by exchange against the sodium Na+ ions present on the resin. The concentration M+ has to be reduced from 15 to 1 mol/m3. The concentration M+ on the incoming regenerated ion exchange resin equals 200 mol/m3. The regeneration liquid contains 3000 mol/m3 Na+ and no M+. a. Determine the equilibrium diagrams for extraction and regeneration when the total ion concentrations in the resin, soil/water mixture and regeneration solution are given by: Resin: [C+] = [Na+] + [M+] = 2400 mol/m3 Soil/water: [C+] = [Na+] + [M+] = 30 mol/m3 Regeneration liquid: [C+] = [Na+] + [M+] = 3000 mol/m3 and the equilibrium follow ideal behavior with an equilibrium constant K = 5. b. Determine the minimal required ion exchanger stream. c. Determine the number of equilibrium stages and concentration M+ in the outgoing ion exchange resin for 1.2 times the minimal ion exchanger stream. d. Determine the minimal value of the regeneration liquid stream to obtain a regenerated resin with 200 mol/m3 residual M+. e. How many equilibrium stages are required with 1.2 times the minimal regeneration stream?

7 Drying of solids

7.1 Introduction Drying is usually defined as the removal of a volatile liquid from solids, slurries or solutions by evaporation to yield solid products. The separation may be carried out in a mechanical manner without phase change or by evaporation through the supply of heat. This thermal drying is discussed in the following sections. Thermal drying consists of two steps. First heat is supplied to the solid material to evaporate the moisture out of the product. Secondly, the vapor is separated from the product phase and, if necessary, condensed outside of the dryer. If possible, mechanical predrying is installed upstream of thermal drying because solids handling is made easier and liquid separation without evaporation is less costly. Fig. 7.1 illustrates a possible position of predrying and thermal drying in an overall process. In industry drying is applied to many types of materials, which differ tremendous in properties and in the required specifications. A few examples are – – – – –

building materials like oxides and clays granular materials like sand and salts food stuff like fruit juices, coffee, milk sheet like materials such as paper, leather fabrics

Also, there are many reasons why a drying step is required. Some examples are: – – – – –

to to to to to

avoid unnecessary transport of water improve the handling properties of a powder preserve food recover remaining traces of a solvent meet the specifications (for storage or as a reaction agent)

Heat may be transferred to the solid in direct dryers by convection, in indirect dryers by conduction and in either case by radiation. The properties of the wet feed and the required specifications of the dried product are of major importance in the choice of the appropriate method of heat transfer and, subsequently the type of dryer. A comprehensive treatment of all kinds of combinations requires more space than available and is thus outside the scope of this textbook. The reader will find these in specialized books such as mentioned in the references and further reading section.

178

Fig. 7.1

7 Drying of solids

Predrying and drying steps in an overall process.

This chapter will be restricted to convective thermal drying of wet solids, mainly by the use of heated air. A common liquid to be removed is water; hence we will often refer to the liquid part as moisture. It should be noted that the theoretical derivations of the fundamental drying equations equally apply to other volatile liquids, except where the use of psychrometric charts is involved, see paragraph 7.6. The content of this chapter is organized along the following lines: – – – –

definitions of specific terminology in drying technology identification of drying mechanisms derivation of simplified rate equations to estimate required drying times drying methods and drying equipment

7.2 Humidity of carrier gas 7.2.1 Definitions Let us consider a stream of heated air at temperature Tf and vapor (water) concentration cf being used to convey heat to a wet surface. After evaporation of liquid, the moisture is subsequently removed as vapor. For low vapor concentrations and low drying rates, the drying rate Φvap is proportional to the driving force csat (Ts) − cf and the rate expression becomes Φvap = kg (csat (Ts) − cf )

[mol s–1 m–2]

(7.1)

7.2 Humidity of carrier gas

where Ts = csat = cf = kg =

temperature of wet surface saturation vapor (water) concentration at Ts vapor (water) concentration in feed gas (air) mass transfer coefficient

179

[K] [mol m–3] [mol m–3] [m s–1]

Two special cases are to be considered. In case the amount of air is large compared to the amount of evaporated moisture, the dynamic equilibrium temperature of the wetted surface is called the wet-bulb temperature Twb . This is the situation existing in a wet-bulb thermometer, where the temperature decrease of the liquid surface balances the heat required for evaporation, see the appendix to this chapter for details. In the stationary situation the drying rate is balanced by the rate of heat conduction from the heated air at Tf to the wetted surface (see also Fig. 7.19 in the appendix to this chapter): [J s–1 m–2]

Φvap ⋅ ΔHvap = h (Tf − Twb) where h = heat transfer coefficient ΔHvap = molar heat of evaporation

(7.2)

[W m–2 K–1] [J mol–1]

Thus, the wet-bulb temperature indicates the maximum amount of vapor that can be carried by an amount of dry gas, the driving force in the heat flux equation being Tf − Twb. On the other hand, when a stream of air Φair at Tf is mixed thoroughly and adiabatically with a lot of liquid, the air leaving the system in equilibrium with that liquid is at saturation temperature Tsat . By definition, the liquid is already at the eventual equilibrium temperature Tsat . By maintaining a constant amount of liquid at Tsat , it does not contribute to the enthalpy balance and the enthalpy decrease of the carrier gas equals just the heat of evaporation: ΔHvap (csat − cf ) Φair = Cp cair Mair (Tf − Tsat) Φair

[J s–1]

or csat − cf =

ρair Cp ΔHvap

(Tf − Tsat)

where Cp = heat capacity of moist air ρair = cair · Mair = density of moist air

(7.3)

[J K–1 (kg dry air)–1 ] [kg dry air (m3 humid air)–1]

The specific volume of moist air is calculated from the volume of a unit weight dry air plus the volume of vapor − at the average temperature (Tf + Tsat)/2 − contained in that amount of air. Therefore, it is convenient to use humidities Hf ,

180

7 Drying of solids

expressing moisture content per unit weight of dry air. The heat capacity Cp of moist air is then easily calculated from that of dry air and that of vapor contained in that air: Cp = Cp, dry air + Hf · Cp, vapor

[J/K.kg]

(7.4)

The moisture content in a gas is directly related to the (water) vapor concentration cf in that gas. The definition of relative humidity Hrel at a given temperature T is: Hrel = 100 ⋅

cf csat (T)

≈ 100 ⋅

pw

[%]

psat (T)

(7.5)

It should be noted that the right-hand side of this equation is accurate only for ideal gases. The absolute humidity Hf is a function of just the vapor concentration at a given total pressure ptot : Hf ≡

cf

⋅

Mw

cair Mair

=

cf

⋅

Mw

ctot − cf Mair

[kg kg/kg dry air]

(7.6)

where Mw represents the molecular weight of (water) vapor. Usually cf ≪ ctot and cair may then be replaced by ctot . For not too high a temperature T, the maximum amount of water per unit weight of dry carrier gas is given by Hsat (T) ≈

csat (T) Mw ctot

(7.7)

Mair

7.2.2 System air-water: a special case According to Eq. 7.1, the drying rate of a completely wetted surface is proportional to the difference in vapor concentration at the LG-interface and that in the vapor phase. Measurement of the wet-bulb temperature provides a means to determine this difference because combination of Eqs 7.1 and 7.2 gives csat (Twb) − cf =

h ΔHvap kg

(Tf − Twb)

(7.8)

Subtraction of Eq. 7.6 from Eq. 7.7 and elimination of csat − cf with Eq. 7.8 leads to a similar expression in terms of humidities1: Hsat − Hf =

1

h

Mw

ρair kg ΔHvap

(Tf − Twb)

(7.9)

It should be noted that in industrial applications the change in enthalpy at evaporation often is expressed per unit weight, the specific heat of evaporation ΔHvap ′ equals ΔHvap / Mw (J/kg).

7.2 Humidity of carrier gas

181

where Hsat should be evaluated at the wet-bulb temperature Twb . This equation applies to any kind of volatile liquid and requires empirical relations to evaluate appropriate values for the mass and heat transfer coefficients, kg and h, respectively. The saturation humidity as a function of adiabatic saturation temperature Tsat follows from Eq. 7.3: Hsat − Hf = Cp

Mw ΔHvap

(Tf − Tsat)

(7.10)

The following reasoning shows that for the system air-water the ratio h/ρair · kg happens to be equal to Cp . Both experimental data on the values of h/kg and theoretical considerations, based on the analogy between mass and heat transfer, show that this ratio is constant. The Chilton-Colburn transfer numbers for heat, jH , and mass, jD , are defined as: jH = Nu Re −1 Pr −1/3 jD = Sh Re −1 Sc −1/3

(7.11)

with Nu =

hd λ

, Sh =

kg d Dg

, Re =

ρair vd η

, Pr =

ν a

=

η ρair Cp ρ

λ

, Sc =

ν Dg

The symbols used here have their usual meaning, for details see a textbook on transport phenomena. The analogy between heat and mass transport for geometrically similar cases and the same flow conditions at not too low a value of Re is j H = jD

(7.12)

Using the definitions above, the ratio of Nusselt and Sherwood is given by Nu Sh

=

hDg kg λ

=

h

1

Pr

kg ρair Cp Sc

=

Pr

−1/3

(Sc )

⇒

h kg

= ρair Cp

Sc 2/3

(Pr)

(7.13)

The ratio h/kg is proportional to the ratio of the thermal and mass diffusivity, which, in a stationary state, only depends on material properties. For air, where h Prandtl and Schmidt are about equal, ≈ Cp and Eq. 7.9 becomes: ρair ⋅ kg Hsat − Hf = Cp

Mw ΔHvap

(Tf − Twb)

(7.14)

Comparing the Eqs. 7.10 and 7.14 reveals that in this special case the wet-bulb temperature equals the adiabatic-saturation temperature. The consequence is that wet bulb temperatures for air-water can be obtained from psychrometric charts.

182

7 Drying of solids

7.2.3 Psychrometric charts The calculation of the humidity Hf in Eq. 7.10 from the experimental wet-bulb temperature Twb requires the saturation humidity Hsat to be known as a function of Twb . The computational efforts involved in this evaluation can be avoided with the help of psychrometric charts. In a psychrometric chart at a given total pressure, the humidity H is plotted as a function of the temperature of a certain gas in a set of curved lines, each representing constant relative humidity. An additional feature of a psychrometric chart is the presence of a set of almost parallel straight lines, adiabatic cooling lines, giving the absolute humidity when gas is adiabatically cooled down. Fig. 7.2 shows the essentials of a psychrometric chart: – a curve representing 100 % relative humidity, the boiling curve at constant total pressure, – an adiabatic cooling line of a carrier gas at temperature Tf and humidity Hf , its slope given by Eq. 7.9, – the (dotted) wet-bulb cooling line of the same gas, its slope given by Eq. 7.8. In the air-water system the adiabatic saturation temperature equals the wet-bulb temperature. In other words, the wet-bulb line and the adiabatic cooling line have the same slope. Thus, the psychrometric chart offers an elegant tool to facilitate the computations involved in solving Eq. 7.10 for Hf . For air-water, the procedure to find the humidity Hf at 1 bar total pressure and temperature Tf is as follows: – Find the wet-bulb temperature on the line with 100 % relative humidity (note that now Twb = Tsat )

Fig. 7.2 Illustration of the main features in a psychrometric chart. In case of the system air-water the two slopes are equal and Tsat = Twb .

7.3 Moisture in solids

183

Fig. 7.3 Psychrometric chart of air-water at 1 bar total pressure (adapted from [56]).

– Follow the adiabatic cooling line until the air temperature Tf is met – Read the humidity Hf on the vertical axis. Reversibly, the wet-bulb temperature Twb – which governs the drying rate according to Eq. 7.2 − can be found from known humidity Hf and feed temperature Tf . Fig. 7.3 gives a psychrometric chart for the system air-water at 1 bar total pressure. All adiabatic cooling lines have their endpoint on the 100 % relative humidity line, where the water vapor in air is in equilibrium with liquid water. There, the saturation humidity Hsat can be read from the vertical axis and the wet-bulb temperature Twb = Tsat from the temperature axis.

7.3 Moisture in solids As the previous section dealt with the properties of moist gases, this section is devoted to moisture in solid materials and focuses on – how is moisture bound in solid particles, and – how does the kind of moisture influence the drying rate

184

7 Drying of solids

7.3.1 Bound and unbound water Many industrially important solids, like activated carbons and molecular sieves, are renowned for their highly porous character, average pore diameters ranging from microns down to molecular sizes. In narrow pores the saturation vapor pressure of a liquid deviates from the equilibrium value for a bulk liquid at the same temperature. The drying rate is proportional to the difference in saturation vapor concentration at the external surface of the wet material and the vapor concentration in the carrier gas (cf Eq 7.1). Thus, to understand the relation between the properties of wet solids and their drying rate, it is necessary to look into the way a volatile liquid is held in the pores of a solid. Two different forms of moisture in solids can be distinguished: bound moisture and unbound moisture. The liquid film at the external surface of solids and the liquid trapped in the spaces between solid particles exert a saturation vapor pressure in the same order of magnitude as that of bulk liquid. In case of water vapor, this kind of water is referred to as unbound or excess water, see Fig. 7.4. All moisture that can be removed by drying is called free moisture which comprises all unbound moisture and part of the bound moisture. The limiting moisture content by drying with a carrier gas at relative humidity H1 is the equilibrium moisture content, Weq . The critical moisture content, Wcrit , separates the two regions with different drying mechanisms, as will be explained in section 7.4. Porous solids also adsorb water in (narrow) pores. In a filled pore, the liquidgas interface is curved and the resulting saturated vapor pressure will be (much) lower than that of the bulk liquid. The more curved the liquid-gas interface, the

Fig. 7.4

Graphical illustration of the various kinds of moisture.

7.3 Moisture in solids

185

lower the saturation pressure. This water will evaporate (much) slower than unbound water. The physical principles underlying the relation between saturated vapor pressure of a liquid in a narrow pore and pore radius will be explained in the following paragraphs. 7.3.2 Saturated vapor pressure in pores The amount of moisture taken up by a porous solid depends on the actual vapor pressure. Generally, the higher the vapor pressure, the higher the uptake. The total pore volume limits the maximum uptake wsat in the solid, see Fig. 7.4. The removal of this moisture, drying, can only be accomplished at a vapor pressure below the saturated vapor pressure of partly or completely filled pores. At the same temperature the saturated vapor pressure psat of a liquid in a pore is smaller ∞ than that of a bulk liquid psat because the liquid-gas (LG-)interface in a pore is concave. To illustrate this effect, in Fig. 7.5a the saturated vapor pressure of a liquid droplet in air and the saturated vapor pressure inside a cavity are compared ∞ of a bulk liquid with a flat interface. The pressure in the liquid phase with psat just below the flat interface, Pliq , equals the total pressure P in the gas phase. Thermodynamic considerations2 show that if for some reason Pliq changes, the saturation pressure psat above the LG-interface changes with Pliq according to ∞ psat = psat exp

Vliq (Pliq − P)

(7.15)

RT

where Vliq = molar volume of liquid. One reason for such a change is a change in the curvature of the LG-interface: in a (very) small droplet (see left hand side in Fig. 7.5a) the liquid pressure is higher to compensate for the force exerted by the surface tension γ. The pressure difference across an LG-interface is given by the Laplace equation Pliq − P = γ

1

(r1

+

1

(7.16)

r 2)

where r1 and r2 are the radii of the curved interface in two directions. The sign of the radius is positive for convex surfaces (see Fig. 7.5a, left hand side) and negative for concave interface (Fig. 7.5a, right hand side). In case of a liquid droplet in the gas phase, r1 = r2 = +rdroplet , whereas in a vapor bubble in liquid r1 = r2 =– rbubble . Application of Eqs. 7.15 and 7.16 to a liquid droplet in a gas phase and to a cavity (vapor bubble) in a liquid phase (see Fig. 7.5a) gives ln

2

psat ∞ psat

=

Vliq γ

⋅

2

R T rdroplet

and

ln

psat ∞ psat

=−

Vliq γ

⋅

2

R T rbubble

(7.17)

See e.g. Atkinsʼ Physical Chemistry, P. W. Atkins and J. de Paula, 7th Edn., Oxford University Press, 2002, Chapter 6.

186

7 Drying of solids

Fig. 7.5a Saturation vapor pressure at curved LG-interfaces. Left: liquid droplet; center : flat interface for comparison; right : bubble in liquid phase.

Fig. 7.5b Saturation vapor pressure of a liquid in a cylindrical pore. Left: filled pore, hemispherical LG-interface, r1 = r2 = − rpore ; right : liquid film, cylindrical interface, r1 = ∞, r2 = − ( rpore − δ).

The increased liquid pressure inside the liquid droplet results in a higher saturated vapor pressure. Analogously, the pressure inside a small vapor bubble in a liquid phase is lower then the liquid pressure. Consequently, the saturation pressure in a cavity is smaller than the equilibrium saturation pressure above a flat interface. The next step is to see what vapor pressures exist in pores. To that purpose we will discuss the filling and drying properties of a hypothetical pore: a straight cylinder with zero contact angle. Such a pore is characterized by two radii: r1 = ∞ in the plane of drawing and r2 = –rpore perpendicular to that plane. Increasing the vapor pressure pvapor in such a pore (see right hand side in Fig. 7.5b) will result a higher amount adsorbed and the average thickness δ of the adsorbed film increases. Hence, the net radius of the pore, rpore − δ, decreases and so does the equilibrium saturated vapor pressure, as prescribed by Eqs. 7.16 and 7.15. As soon as pvapor > psat , or ∞ pvapor ≥ psat exp −

Vliq

γ

( R T rpore − δ)

(7.18)

7.3 Moisture in solids

187

Fig. 7.6 Water adsorption in porous solids. (a) Hypothetical porous solid with straight cylindrical pores of equal diameter; (b) Adsorption-desorption branches of irregular shaped pores.

the pore will be completely filled with liquid, instantaneously. The expression in Eq. 7.18 is known as the Kelvin-equation for a partly filled straight cylinder. This liquid-vapor equilibrium is not reversible. A slight decrease in pvapor will not empty the pore because the LG-interface is now hemispherical with radii equal to rpore , see left hand side in Fig. 7.5b. This interface is stronger curved than the interface in the partly filled pore. Consequently, the requirement for emptying the pore is more severe: pvapor < psat , or ∞ pvapor < psat exp −

Vliq 2γ

( R T rpore)

(7.19)

which is the form the Kelvin equation takes in case of a filled cylindrical pore. The phenomenon that emptying a pore occurs at a lower vapor pressure than its filling is known as hysteresis and shows up as separated adsorption and desorption branches in adsorption isotherms as those presented in Fig. 7.6. An important feature of the adsorption isotherm of a hypothetical solid, containing straight cylindrical pores of equal size, can be deducted. At low vapor pressure a thin film will be present on the inside of the pores. At increasing vapor pressure this adsorbed film will grow in thickness (following an isotherm equation relating vapor pressure and film thickness). As soon as the vapor pressure reaches the limiting value given by Eq. 7.18, all pores will be completely filled instantaneously. This behavior is depicted in the right-hand branch of Fig. 7.6a3. Reversely, starting from complete saturation, the moisture content with decreasing vapor pressure will follow the left-hand branch in Fig. 7.6a. All pores are 3

Note that, unlike the plot in Fig. 7.4, where moisture content is plotted as a function of relative humidity, this isothermal graph gives moisture content as a function of vapor pressure.

188

7 Drying of solids

partially emptied as soon as the vapor pressure is below the limiting value indicated by Eq. 7.19. In contrast with the hypothetical adsorption isotherm in Fig. 7.6a, a real adsorption isotherm, of water on TiO2 with a distribution of pore forms and sizes is shown in Fig. 7.6b. These examples illustrate that the saturation pressure required for drying should be below the closing point pdes of the hysterisis loop to ensure a satisfactorily driving force for drying.

7.4 Drying mechanisms In the design of drying operations an understanding of liquid and vapor mass transfer mechanisms is essential for quality control. Because no two materials behave alike, this understanding is usually obtained by measuring drying behavior under controlled conditions in a prototypic, pilot-plant dryer. From these experiments the drying rate is usually determined as the change of moisture content with time. For most moist solids, especially those having capillary porosity, the drying rate depends upon the moisture content in a manner similar to that shown in Fig. 7.7a, which gives moisture content as a function of time, or, related, Fig. 7.7b, showing the drying rate as a function of moisture content. Three regions can be observed. In the initial drying period AB the liquid in the solid has a lower temperature and it needs some time to get to its stationary drying state. This initial heating period is usually quite short and can often be ignored. In the region BC the drying rate is constant, whereas in the region CE the drying rate is decreasing. Point E represents the lowest possible moisture content of the solid, Weq in Figs. 7.4 and 7.7 at the specified moisture content of the gas.

Fig. 7.7

(a) Drying curve; (b) Drying rate curve.

7.4 Drying mechanisms

189

7.4.1 Constant drying rate Initially, the surface of the solid is initially very wet and a continuous film of water exists on the drying surface. This water is entirely unbound and acts as if the solid were not present. Under these conditions the rate of evaporation is independent of the solid and is essentially the same as the rate from a free liquid surface. The evaporation rate is controlled by the heat-transfer rate to the wet surface. In the region AB (Fig. 7.7) the mass-transfer rate adjusts to the heat transfer rate and the wet surface reaches the wet-bulb temperature Twb . After reaching this stationary state temperature the drying rate remains practically constant for a period of time, region BC. The resulting constant drying rate is obtained from Eq. 7.1 rC = kg (csat (Twb) − cf)

[kg s–1 m–2]

(7.20a)

[kg s–1 m–2]

(7.20b)

or, from Eq. 7.2, rC =

h (Tf − Twb) ΔHvap Mw

where ΔHvap = molar heat of evaporation [J mol–1] and Mw the molar weight of (water) vapor (cf. note on page 180). Given the driving force, the constant drying rate rC in this first period depends on the mass transfer coefficient kg and the heat transfer coefficient h (cf. Eq. 7.20) and thus on the velocity v of the carrier gas. Turbulent flow parallel to a drying surface will cause more turbulence then a well-developed flow parallel to a flat plate. Thus, the usual Sherwood and Nusselt relations predict too low a value for the transfer coefficients. An educated guess for this effect and applying average values for the physical constants of the air-water system at the temperature interval of interest (say 40 to 140 °C) lead to the following expression for the heat transfer coefficient: h = 12 (ρv)0.8

(7.21)

with ρ = density of moist air [kg moist air / unit volume of moist air] and v = velocity of moist air [m s–1]. The drying rate remains constant as long as the external conditions are constant. Under these conditions all principles relating to simultaneous heat and mass transfer between gases and liquids apply. If the solid is porous, most of the water evaporated in the constant rate period − free moisture as defined in Fig. 7.4 − is supplied from the interior of the solid. The constant rate period continues only as long as the water is supplied to the surface as fast as it is evaporated.

190

7 Drying of solids

7.4.2 Falling drying rate When the moisture content is reduced below a critical value, Wcr in Figs. 7.4 and 7.7, the surface of the solid dries out, and a further decrease of the free moisture content starts taking place in the interior of the porous solid. This is called the falling drying rate period. In this period, the velocity of the carrier gas is less important due to the transport of the moisture from within the solid to the outer surface. First, the wetted surface area decreases continuously until the surface is completely dry and the plane of evaporation slowly recedes from the surface. Heat for evaporation is transferred through the solid to the zone of evaporation. Then the drying rate is controlled by the internal material moisture transport and decreases with decreasing moisture content. Liquid diffusion may become rate determining in the first part of this period and then the moisture content at the external surface is at the equilibrium value Weq . Fickʼs second law for unsteady-state diffusion is applicable, but is difficult to solve because the liquid diffusivity varies with moisture content and humidity. Like in other diffusion problems, however, the general picture is still that the drying rate is directly proportional to the moisture content and inversely proportional to the square of the thickness of the drying layer. In wet layers of small, granular particles (e.g. clays, minerals), water may get to the surface by capillary flow rather then by liquid diffusion. Now the drying rate is determined by both capillary flow and evaporation from the surface. Theoretical considerations (outside the scope of this chapter, but the reader is referred to Geankoplis [1] for elucidation of this point) lead to a rate model showing that the drying rate is inversely proportional to the thickness of the drying layer. The mechanism of evaporation is still the same as in the constant-rate stage as long as a continuous liquid film across the pores is maintained, so gas velocity, humidity and temperature of the carrier gas also influence the overall drying rate. Therefore, finding experimental evidence for either mechanism is not too difficult. The amount of moisture removed in the falling rate period may be relatively small but the required time is usually long. At the end of the falling rate period the residual moisture in the solid is bound by sorption. Heat conduction through the solid and diffusion of vapor in the pores become rate determining. The drying rate decreases rapidly with decreasing moisture content and tends to zero as the (hygroscopic) equilibrium moisture content Weq is approached. 7.4.3 Estimation of drying time The form of the drying rate curve suggests that the determination the time necessary to dry a solid until the required moister content is obtained, should be carried out separately for the first drying period (BC in Fig. 7.7b) and for the second drying period (CD in Fig. 7.7b). Because there is no way to predict the critical

7.4 Drying mechanisms

191

moisture content, which separates the two regimes, an experimental drying rate curve is required. Defining the drying rate rC in terms of the change of moisture content W with time: rC = −

w0 dW

[kg s–1 m–2]

A dt

where w0 = weight of dry solid A = evaporating surface area

(7.22)

[kg] [m2]

the time tC for the first drying period is obtained by integration of Eq. 7.22 between the moisture content of the wet feed, W0 , and the critical moisture content Wcr : tC =

w0 W0 − Wcr A

[s]

rC

(7.23)

The time necessary for drying from the critical moisture content Wcr until the desired moisture content Wfinal4 is given by: Wfinal

tF = −

w0 A

∫

dW

[s]

rF

Wcr

(7.24)

One way to develop the integral is to plot the reciprocal values of rF as a function of the moisture content and to subsequently determine the area under this curve and between the two boundaries. On the other hand, an analytical solution can be obtained when the dependence of rF with moisture content is known. The shape of the falling drying period curve depends, partly, on the type of material to be dried. Some porous materials with relatively large external surface, like sheets of paper, show a linear change of drying rate with moisture content, see Fig. 7.8: rF = a + b ⋅ W

(7.25)

Substitution in Eq. 7.24 gives: tF =

w0 b⋅A

ln

a + b ⋅ Wcr a + b ⋅ Wfinal

=

w0 b⋅A

ln

rC rF, final

(7.26)

Also slope b in this equation can be eliminated, see Fig. 7.8: tF = 4

Wcr − Wfinal w0 rC − rF, final A

ln

rC rF, final

The drying rate becomes zero approaching the equilibrium content Weq , hence Wfinal > Weq

(7.27)

192

Fig. 7.8

7 Drying of solids

Elimination of slope b in Eq. 7.26 from drying rate curve

rC − rF, final Introducing the mean logarithmic drying rate rF, lm = , Eq. 7.26 rC becomes ln rF, final w0 Wcr − Wfinal tF = (7.28) A rF, lm In either case the total drying time ttot simply follows from ttot = tC + tF =

w0 W0 − Wcr A[

rC

+

Wcr − Wfinal rF, lm

]

(7.29)

7.5 Classification of drying operations Drying methods and processes can be classified in several ways. One classification is as batch, where the material is inserted into the drying equipment and drying proceeds for a given period of time, or continuous, where the material is continuously added to the dryer and dried material continuously removed. A batch dryer is best suited for small lots and for use in single-product plants. This dryer is one into which a charge is placed, the dryer runs through its cycle, and the charge is removed. In contrast, continuous dryers operate best under steady-state conditions drying continuous feed and product streams. Optimum

7.5 Classification of drying operations

193

Fig. 7.9 Convective drying in closed system.

operation of most continuous dryers is at design rate and steady state. Continuous dryers are unsuitable for short operating runs in multiproduct plants. Drying processes can also be classified to the physical conditions used to add heat and remove water vapor. The most frequently applied heat transfer mechanisms are convection drying and contact drying. In convection drying the sensitive heat of a hot gas is supplied to the material surface by convection. The drying agent flowing past or through the body also removes the evaporated water and transports it from the dryer (Fig. 7.1). For energy saving, partial recirculation of the drying medium is also used. Drying operations involving toxic, noxious, or flammable vapors employ gas-tight equipment combined with recirculating inert gas systems having integral dust collectors, vapor condensers and gas preheaters (Fig. 7.9). In contact drying the heat is supplied to the wet material by conduction from the heated surface as bands, plates, cylinders or the dryer wall. The amount of heat transferred depends not only on the thermal conductivity of the heating surface but also on the heat transfer coefficient from the heating medium to the surface. Common heating media include steam, organic liquid and molten metals. Since all the heat for moisture evaporation passes through the material layer the thermal efficiency of contact drying is higher than convective drying where most of the heat is flowing over the material and wasted into the outlet air. During constant rate drying, material temperature is controlled more easily in a directheat dryer than in an indirect-heat dryer because in the former the material temperature does not exceed the gas wet bulb temperature as long as all surfaces are wet.

194

7 Drying of solids

7.5.1 Direct-heat dryers In direct-heat dryers, steam heated, extended surface coils are used to heat the drying gas up to temperatures as high as 200 °C. Electric and hot oil heaters are used for higher temperatures. Diluted combustion products are suitable for all temperatures. In most direct-heat dryers, more gas is needed to transport heat than to purge vapor. Some of the most commonly used direct-heat dryers are listed below. Batch compartment dryers Direct heat batch compartment dryers are often called tray dryers because of frequent use for drying materials loaded in trays on trucks or shelves. Fig. 7.10 illustrates a two-truck dryer. The compartment enclosure comprises insulated panels designed to limit exterior surface temperature to less than 50 °C. Slurries, filter cakes, and particulate solids are placed in stacks of trays. Large objects are placed on shelves or stacked in piles. Unless the material is dusty, gas is recirculated through an internal heater as shown. Only enough purge is exchanged so as to maintain needed internal humidity. For inert gas operation, purge gas is sent through an external dehumidifier and returned. These dryers are economical only for single-product rates less than 500 t/yr, multiple product operation, and batch processing. Variable speed fans are employed to provide higher gas velocity over the material during early drying stages. To minimize dusting, the fans reduce velocity after constant drying when heat transfer at the material surface is no longer the limiting drying mechanism. Shallow tray loading yields faster drying, but care is needed to ensure depth uniformity and labor is increased.

Fig. 7.10

Two-truck tray dryer (adapted from [17]).

7.5 Classification of drying operations

195

Fig. 7.11 Radiation belt dryer.

Belt dryers In belt dryers, such as shown in Fig. 7.11, a loading device that is especially designed for the product is used to place the moist solid on the surface of a circulating belt, which passes through a drying chamber that resembles a tunnel. The solid remains undisturbed while it dries. At the end of this chamber the material falls from the belt into a chute for further processing. In some installations the material falls onto another belt that moves in the opposite direction to the first one. Centrifugal or axial flow blowers are used to aerate the moist materials. The air stream can enter the solid from below or above. Belt dryers are ideal for friable, molded, granular, or crystalline products that require a long, undisturbed drying time. They are used in all branches of industry. Rotary dryers The direct-heat rotary dryer in Fig. 7.12 is a horizontal rotating cylinder through which gas is blown to dry material that is showered inside. Shell diameters are 0.5–6 m. Batch rotary dryers are usually one or two diameters long. Continuous dryers are at least four and sometimes ten diameters long. At each end, a stationary hood is joined to the cylinder by a rotating seal. These hoods carry the inlet and exit gas connections and the feed and product conveyors. For continuous drying the cylinder may be slightly inclined to the horizontal to control material flow. An array of material showering flights of various shapes is attached to the inside of the cylinder. Dry product may be recycled for feed conditioning if material is too fluid or sticky initially for adequate showering. Slurries may also be sprayed into the shell in a manner that the feed strikes and mixes with a moving bed of dry particles. Material fillage in a continuous dryer is 10–18 % of cylinder volume. Greater fillage is not showered properly and tends to flush toward the discharge end.

196

7 Drying of solids

Fig. 7.12

Direct heat rotary dryer.

Direct-heat rotary dryers are the workhorses of industry. Most particulate materials can somehow be processed through them. These dryers provide reasonably good gas contacting, positive material conveying without serious back-mixing, good thermal efficiency, and good flexibility for control of gas velocity and material residence time. Gas flow in these rotary dryers may be cocurrent or countercurrent. Cocurrent operation is preferred for heat-sensitive materials because gas and product leave at the same temperature. Countercurrent allows a product temperature higher than the exit gas temperature, which increases the dryer efficiency. To prevent dust and vapor escape at the cylinder seals, most rotary dryers operate at a slightly negative internal pressure. Flash dryers In flash dryers materials are simultaneously transported and dried. The simplest form, illustrated in Fig. 7.13, consists of a vertical tube in which granular or pulverized materials are dried while suspended in a gas or air stream. The available drying time is only a few seconds. Only fine materials with high rates of heat and mass transfer or coarser products with only surface moisture to be removed are used in such dryers. Solids that contain internal moisture can only be dried to a limited extend by this method. Flash dryers are well suited for drying thermally sensitive materials and are widely used for drying organic and inorganic salts, plastic powders, granules, foodstuffs etc. Fluidized bed dryers These units are known for their high drying efficiency. A two-stage model is illustrated in Fig. 7.14. The solid moves horizontally in a chute and the drying

7.5 Classification of drying operations

Fig. 7.13

Flash dryer.

Fig. 7.14

Two-stage fluid bed dryer.

197

agent flows vertically through a perforated floor to fluidize the solid. These machines can operate continuously, because the solid is transported while suspended in the drying agent. Materials that can be suspended in the drying agent are usually powders, crystals and granular or short-fibred products that remain finely

198

7 Drying of solids

divided. Pastes and slurries are mixed with previously dried material and are then easily fluidized. Spray dryers Spray drying is used for the drying of pastes, suspensions or solutions. The moist material is sprayed into the drying agent and converted into a powder that is entrained by the gas stream. A spray dryer is a large, usually vertical chamber through which hot gas is blown and into which a solution, slurry, or pump able paste is sprayed by a suitable atomizer. The largest spray-dried particle is about 1 mm, the smallest about 5 μm. Because all drops must reach a non-sticky state before striking a chamber wall, the largest drop produced determines the size of drying chamber. Chamber shape is determined by nozzle or disk spray pattern. Nozzle chambers are tall towers, usually having height/diameter ratios of 4–5. Disk chambers are of large diameter and short. A spray dryer may be cocurrent, countercurrent, or mixed flow. Cocurrent dryers are used for heat-sensitive materials. Countercurrent spray dryers yield higher bulk density products and minimize hollow particle production. Fig. 7.15 shows an open-cycle, cocurrent, disk atomizer chamber with a pneumatic conveyor following for product cooling. Spray dryers are often followed by fluid beds for second-stage drying or fines agglomeration. Spray dryers are particularly suitable for drying solids that are temperature sensitive. Applications include coffee and milk powders, detergents, instant foods, pigments, dyes etc.

Fig. 7.15

Spray dryer system (adapted from [17]).

7.5 Classification of drying operations

199

7.5.2 Contact dryers In contact dryers most of the heat is transferred by conduction. Common heating media are steam and hot water, depending on the actual operating temperature. Based on dryer cost alone, contact dryers are more expensive to built and install than direct heat dryers. For environmental concerns contact dryers are more attractive because they are more energy efficient and use only small amounts of purge gas. Dust and vapor recovery systems for contact dryers are smaller and less costly. In this paragraph some of the commonly encountered contact dryers are described. Rotary and agitator dryers The heat necessary for drying is transferred through the peripheral walls of these dryers in contact drying. Only a small amount of air is necessary to carry of the moisture that is taken from the solid. Accordingly the air velocity in these units is quite low. This is advantageous when drying materials that dust easily or form dust during drying. The rotary steam tube-dryer is a horizontal rotating cylinder in which one or more circumferential rows of steam-heated tubes are installed. In agitator dryers (Fig. 7.16) the vessel is stationary and the solids product is slowly agitated. Material holdup may be varied from a few minutes to several hours. Agitator speeds rarely exceed 10 min–1, because at higher speeds the mechanical stress and power demand become intolerable.

Fig. 7.16

Indirect heat paddle-type agitator dryer.

Vacuum dryers The principal differences in the vacuum dryers are their seals and the means to produce the vacuum. Drying under reduced pressure is advantageous for materials that are temperature sensitive or easily decomposed because the evaporation tem-

200

7 Drying of solids

Fig. 7.17 Tumble vacuum dryer.

perature is reduced. They are most often used to dry pharmaceutical products and foodstuffs. The simplest form of a vacuum dryer for batch drying is the vacuum shelf dryer. The moist solid lies on a heated plate. Improved heat transfer with higher efficiency is obtained in the vacuum tumble dryer of Fig. 7.17, in which the moist solid is constantly agitated and mixed. These tumble dryers are also used for the final drying step during the production of various polyamides. Fluid bed dryers Indirect-heat fluid bed dryers are usually rectangular vessels in which vertical pipe or plate coils are installed. Fig. 7.18 is diagram of a two-stage, indirect-heat

Fig. 7.18 Indirect heat fluid bed dryer.

7.5 Classification of drying operations

201

fluid bed incorporating pipe or plate coils heaters. The general design is used for several particulate polymers. Excellent heat transfer is obtained in an environment of intense particle agitation and mixing. Because of its favorable heat- and masstransfer capabilities, flexibility for staging and lack of rotary seals, an indirectheat fluid bed is an ideal vessel for vapor recovery and drying in special atmospheres. 7.5.3 Other drying methods Some other, less frequently used drying methods are radiation heating, dielectric drying, and freeze-drying. In radiation heating the energy is supplied from an electromagnetic radiation source that is remote from the surface of the solid. Vibrations of molecules in the wet product, creating the thermal effect and evaporating moisture, absorb the electromagnetic radiation energy. Radiation drying is expensive because of the high cost of electrical power or combustion gas to heat the radiators. Since penetration depth of infrared waves is relatively small, it is mainly used for short drying times of thin product layers such as films, coatings and paints. With high frequency or dielectric drying, the wet product forms a dielectric between the electrodes of a plate capacitor exposed to a high frequency electric field. The generated intermolecular friction causes the generation of heat inside the product. This is used to heat up the product and to evaporate the moisture. High frequency drying allows gentle thermal drying of the product. Substantial deformation and shrinkage cracks are avoided. Therefore, dielectric drying is employed for gentle drying of products such as fine wood, ceramic products, foods, pharmaceuticals and luxury goods. Freeze-drying is a vacuum sublimation drying process. At temperatures below 0 °C and under vacuum, moisture sublimes from a frozen wet product directly from the solid to the gaseous state. Firstly the wet product has to be frozen from –15 to approximately –50 °C. The freezing rate and final temperature essentially determine the drying time and final quality of the product such as structure, consistency, color and flavor. The frozen product is usually granulated, sieved and then charged to the dryer. Under vacuum it is then dried either discontinuously on heated plates or continuously while mixed and moved over a heated surface. Sublimed moisture vapor desublimes to ice on a cooling agent operated condenser. Drying rates are low because the low allowable rate of heat flow controls the process. Freeze or sublimation drying is carried out in a chamber freezer, vacuum disc dryer, vibrating film dryer, cascade dryer or spray freezing dryer. The high investment and operating costs of freeze-drying are only worthwhile for high-grade, thermally sensitive products. Certain important properties of the products are kept such as flavor, taste and color.

202

7 Drying of solids

7.6 Experimental determination of moisture content in air from wet bulb temperature A liquid in equilibrium with its surrounding gas phase exerts the saturation pressure at the particular temperature of that liquid. A gas stream flowing along the surface causes evaporation taking place if and as long as the actual vapor pressure is below the saturation pressure at that temperature. The evaporation from the liquid surface results in a temperature decrease in a thin liquid surface layer. As a result, heat will be transported from the gas to the liquid phase. Fig. 7.19 shows the combined heat and mass transport. The change in molal enthalpy connected to the phase change is then balanced by heat transport from the gas to the liquid. The surface temperature Ts initially drops, decreasing Φvap and increasing Φh , until the stationary state is established. The so-called wet-bulb temperature Twb is the steady state temperature Ts , which is eventually reached when a gas is flowing at constant velocity along a wetted surface. The higher the moisture content in this gas, the smaller the driving force for evaporation and the smaller the difference between the wet-bulb temperature and the temperature of the fluid. By definition, the wet-bulb temperature in a saturated gas is the same as the fluid temperature Tf . In Fig. 7.20 the experimental set-up is shown. The feed gas with unknown humidity Hf is entering at temperature Tf , also designated dry-bulb temperature,

Fig. 7.19 Combined heat and mass transfer.

Fig. 7.20

Device for the measurement of wet-bulb temperatures.

7.6 Experimental determination of moisture content in air

Fig. 7.21

203

Determination of air humidity from wet bulb temperature.

which is measured with the first sensor. The second temperature sensor is positioned in a cylinder in such a way that no radial heat transfer resistance occurs between the sensor and the outer surface: the temperature in the cylinder is uniform. The outer surface is provided with a wick, which is kept wet, any water lost by evaporation from the surface is replaced with water at temperature Tf from a storage vessel connected to the wet wick. No conduction of heat is supposed to occur along the temperature sensors. For the system air-water, in section 7.2.2 the following relation between wet bulb temperature and humidity Hf has been derived: Hsat − Hf = Cp ⋅

Mw ΔHvap

(Tf − Twb)

(7.14)

This is the basic equation to calculate unknown moisture contents Hf from measured wet-bulb temperatures Twb . The graphical solution of this equation is shown Fig. 7.21. Humidities can be converted to concentrations or partial pressures with Eqs. 7.5–7.7. Nomenclature a A c Cp Dg d H Hrel

thermal diffusivity (λ /ρ ⋅ Cp) in Eq. 7.11 surface area concentration heat capacity (of dry air) coefficient of diffusion (characteristic) diameter humidity relative humidity

[m2s–1] [m2] [molm–3] [JK–1(kg dry)–1] [m2s–1] [m] [kg(kg dry)–1] [%]

204

7 Drying of solids

ΔHvap h kg M p R r r T V v w W δ γ λ η ρ Φ

molar heat of evaporation heat transfer coefficient mass transfer coefficient in gas phase molar weight pressure gas constant specific rate radius temperature volume (superficial) velocity weight moisture content (film) thickness surface tension thermal conductivity kinematic viscosity density molar flux

[Jmol–1] [Wm–2K–1] [ms–1] [kgmol–1] [Pa] [Jmol–1K–1] [kg s–1m–2] [m] [K] [m3] [ms–1] [kg] [kg(kg dry)–1] [m] [Nm–1] [Wm–1K–1] [m2s–1] [kgm–3] [mols–1m–2]

Indices m f s sat h vap wb w C F des ads

molar feed surface saturation/saturated heat evaporation/evaporated wet bulb water constant (drying rate) falling (drying rate) desorption adsorption

Exercises 1. Air of 20 °C and 1 bar with a relative humidity of 80 % is heated to 50 °C. The heated air flows along a flat tray filled with completely wetted material. The heat transfer coefficient h = 35 W/m2K and the heat of evaporation ΔHv = 2.45 106 J kg–1. Calculate rc , the constant drying rate. 2. A wet solid with a moisture content W = 0.5 kg kg–1 dry occupies 0.1 m2 on a flat tray. A laboratory experiment shows that the critical moisture content Wc = 0.2 kg kg–1 dry and the constant drying rate rc = 0.36 g s–1 m–2. How long will it take, under similar conditions, to reduce the moisture content of the wet solid on the tray, containing 2 kg solid (dry basis), from 0.4 to 0.2 kg kg–1 dry? 3. A wet solid is dried with air at 60 °C and a humidity Hf = 0.010 kg H2O/kg dry air. The air flows parallel to the flat, wet surface of the solid at a velocity v = 5 m s–1. Under these conditions the heat transfer coefficient h is given by h = 14.3 (ρv)0.8,

Exercises

205

where ρ is the density of the moist air in kg moist air per unit volume moist air. The surface area available for heat exchange with air is 0.25 m2 and the heat of evaporation ΔHv = 2.45 106 J kg–1. Calculate the total amount of water evaporating per unit time. 4. Calculate the relative change in saturation vapor pressure of water at 300 K when it is dispersed in droplets of 0.1 mm radius. γw = 0.073 N m–1. 5. Air at 20 °C and 1 bar with a relative humidity of 80 % is available for drying a porous solid containing cylindrical pores with a diameter of 0.8 nm. The heat of evaporation ΔHv = 44045 J mol–1. To what temperature should the air be heated at least to avoid that the pores remain filled with water? (Hint: the saturated vapor pressure obeys Clapeyron) 6. Wet air (0.2 kg/s) of 30 °C and 2 bar with a relative humidity (Hrel = pwater / p0water) of 75 % has to be dried by adsorption of water on porous alumina in a packed bed. The adsorption isotherm is given by Henryʼs law: kg water adsorbed kg dry alumina

= K ⋅ Hrel

with K (30 °C) = 0.350 and K (60 °C) = 0.121. The process is conducted in an adsorption vessel with a length of 4 m and diameter of 0.5 m. Adsorption and regeneration take place at a constant temperature. The bed porosity εbed = 0.32 for a complete filling with the alumina particles of 2 mm diameter. The particle porosity εparticle = 0.58. At the start of the drying process the water vapor pressure increases linearly until the maximum value of 75 % relative humidity is reached. Furthermore it is given that the saturated vapor pressure of water, p0water equals 0.0418 bar at 30 °C and 0.199 bar at 60 °C. Mw = 0.018 and Mair = 0.029 kg/mol. a. Calculate the maximum weight of dry alumina the adsorption vessel can contain. b. When the first trace of water is detected at the column exit, only 70 % of the column is used for adsorption. Draw the concentration profile in the adsorption vessel (relative humidity as function of the place). c. How long does it take to achieve this situation of 70 % column utilization? d. Calculate the amount of water adsorbed to the alumina. e. How long does it take to completely regenerate the column by heating the inlet air to 60 °C and reducing the pressure to 1 bar.

8 Crystallization and precipitation

8.1 Introduction Crystallization is one of the oldest unit operations in the portfolio of industrial separations. Large quantities of crystalline products are manufactured commercially such as sodium chloride, sucrose, and fertilizer chemicals (ammonium nitrate, ammonium phosphates, urea). In the production of organic chemicals, crystallization is used to recover product, to refine intermediate chemicals and to remove undesired salts. Crystalline products coming from the pharmaceutical, fine chemical and dye industries are produced in relatively small quantities but represent a valuable and important industrial sector. Crystallization distinguishes itself from other separation technologies in that a solid phase is generated. In the process the feed consists of a homogeneous solution that is made supersaturated by cooling and/or evaporation of the solvent to obtain the desired solute in a solid form. The process of solid phase formation is termed crystallization, and the operation occurs in a vessel called a crystallizer. Functions that can be achieved by crystallization include separation, purification, concentration, solidification and analysis. Crystallization usually occurs on added seed crystals or on very small crystals, nuclei, broken off of other crystals. The crystal particles grow and become larger until they are removed (harvested) from the crystallizer. Although the crystals are often of high purity, their performance is also characterized by its inherent shape and size distribution. These solid phase properties are especially important because crystallization is frequently the initial step in a solid processing sequence, similar to that shown in Fig. 8.1. After crystallization the solids are normally separated from the crystallizer liquid, washed and discharged to downstream drying equipment. Since product size and suspended solids concentration are controlled to a large extend in the crystallizer, predictable and reliable crystallizer performance is essential for smooth operation of the downstream system. The final design of a crystallizer is the culmination of the design strategy depicted in Fig. 8.2. In the conceptual design stage (paragraph 8.3), equilibrium data and the operating mode (method of supersaturating generation) are surveyed. Solvent choice and processing conditions are determined in this step. After establishing the type of crystallizer, finally the crystallizer functional design is made (paragraph 8.4). The use of mathematical models for the design of industrial crystallizers has lagged behind compared to other unit operations because of the complexities associated with rationally describing and interrelating growth and

208

8 Crystallization and precipitation

Fig. 8.1 Schematic of a crystallization initiated solids processing sequence.

Fig. 8.2 Crystallization system design strategy.

nucleation kinetics with the process configuration and mechanical features of the crystallizer. The design and analysis of crystallization processes of the continuous well-mixed suspension type have developed into formal design algorithms, which can now be applied in situations of industrial importance. Examples of specific process configurations that can be modeled rigorously include fines destruction, clear-liquor advance, classified-product removal, vessel staging and seeding. The basic requirement for these rigorous process configuration models is a population balance crystal size distribution algorithm.

8.2 Crystal characteristics

209

The main advantages of crystallization are that nearly pure solid product can be recovered in the desired shape in one separation stage. With a careful design, product purity in excess of 99.0 % can be attained in a single crystallization, separation and washing sequence. This large separation factor is one of the reasons that make crystallization a desirable separation method. A second reason why crystallization if often the separation method of choice is that it can produce uniform crystals of the desired shape. During crystallization the conditions are controlled so that the crystals have the desired physical form for direct packaging and sale. This is important when a solid product is desired. The major disadvantages of crystallization are that purification of more than one component and/or full solute recovery are not attainable in one stage. Thus, additional equipment is required to remove the solute completely from the remaining crystallizer solution. Since crystallization involves processing and handling of a solid phase, the operation is normally applied when no alternative separation technique is discernable. Some important considerations that may favor the choice of crystallization as the preferred option over other separation techniques such as distillation include: • Solute is heat sensitive and/or a high boiler that decomposes at the high temperatures required to conduct distillation • Low relative volatility between the solute and contaminants and/or the existence of azeotropes between solute and contaminants • Solid solute product. For example, after purification via distillation a solute must be solidified by flaking or prilling and crystallization may be more convenient to employ. • Comparative economics favor crystallization. If distillation requires high temperatures and energy usage, crystallization may offer economic incentives. Precipitation is an important separation method in the production of many fine chemicals and in mineral processing. It is a related process since solutes dissolved in a solvent precipitate out. However, the precipitate is usually amorphous and will have a poorly defined shape and size. Precipitates are often aggregates of several species and may include salts or occluded solvent. Thus precipitation serves as “rough cut” to either remove impurities or to concentrate and partially purify the product.

8.2 Crystal characteristics In crystallization processes product requirements are a key issue in determining the ultimate success in fulfilling the function of the operation. Key parameters of product quality are the size distribution (including mean and spread), the morphology (including habit or shape and form) and purity. Crystal size distribution (CSD) determines several important processing and product properties. It is often important to control the CSD. Most favored is monodisperse where all crystals

210

8 Crystallization and precipitation

are of the same size and dissolve at a known and reproducible rate. Critical phenomena that influence the CSD outside nucleation and growth are breakage and agglomeration. Breakage of crystals is almost always undesirable because it is detrimental to crystal appearance and it can lead to excessive fines and have a deleterious effect on crystal purity. Agglomeration is the formation of a larger particle through two or more smaller particles sticking together. 8.2.1 Morphology Every chemical compound has a unique crystal shape that depends enormously on the conditions in the crystallizer. Some substances illustrate polymorphism meaning that the substance can crystallize into tow or more unique forms. A good example is carbon, which can crystallize into graphite and diamond. The general shape of a crystal is referred to as habit. A characteristic crystal shape results from the regular internal structure of the solid with crystal surfaces forming parallel to planes formed by the constituent units. The unique aspect of the crystal is that the angles between adjacent faces are constant. The surfaces (faces) of a crystal may exhibit varying degrees of development but not the angles. With constant angles but different sizes of the faces the shape of a crystal can vary enormously. This is illustrated in Fig. 8.3 for three hexagonal crystals. The final crystal shape is determined by the relative growth rates of the crystal faces. Faster growing faces become smaller than slower growing faces and may disappear from the crystal altogether in the extreme. The relative growth process of faces and thereby final crystal shape are affected by many variables such as rate of crystallization, impurities, agitation, solvent used, degree of supersaturation and so forth. Especially the presence of impurities can alter the growth rates of crystalline materials significantly. Most common is a decrease in growth rate, which is considered to involve adsorption of the impurity onto the crystal surface. Once located on the surface, the impurity forms a barrier to solute transfer from the solution to the

Fig. 8.3 Various shapes of a hexagonal crystal.

8.2 Crystal characteristics

211

crystal. Another important effect associated with the presence of impurities is that they may change the crystal shape (habit). Habit alteration affects the appearance of the crystalline product, purity and its processing characteristics (such as washing and filtration) and is considered to result form unequal changes in the growth rates of different crystal faces. The effect on purity becomes even more important when crystallization is employed as a purification technique. Mechanisms by which impurities can be incorporated into crystalline products include adsorption on the crystal surface, solvent entrapment in cracks, crevices and agglomerates and inclusion of pockets of liquid. An impurity having a structure sufficiently similar to the material being crystallized can also be incorporated into the crystal lattice by substitution or entrapment.

8.2.2 Crystal size distribution Particles produced by crystallization have a distribution of sizes that varies in a definite way over a specific size range (Fig. 8.4). A crystal size distribution (CSD) is most commonly expressed as a population (number) distribution relating the number of crystals at each size to the size or as a mass (weight) distribution expressing how mass is distributed over the size range. The two distributions are related and affect many aspects of crystal processing and properties. An average crystal size can be used to characterize a crystal size distribution. However, the average can be determined on any of several bases such as number, volume, weight, and length. The dominant crystal size LD is most often used as a representation of the product size. The coefficient of variation (cv) of a distribution is a measure of the spread of the distribution about the characteristic size. It is often

Fig. 8.4 Crystal size distribution.

212

8 Crystallization and precipitation

used in conjunction with dominant size to characterize crystal populations through the equation: cv =

σ

(8.1)

LD

where σ is the standard deviation of the distribution.

8.3 Crystallization from solutions 8.3.1 Operating modes The technique employed to generate supersaturation in a solution is referred to as the mode of operation. The mode chosen by the designer is strongly influenced by the phase-equilibrium characteristics of the system, and it dictates the material and energy balance requirements of the system. Therefore accurate solid-liquid equilibrium data are essential to evaluate the process design options for crystallization processes. A saturated solution is a solution that is in thermodynamic equilibrium with the solid phase of its solute at a specified temperature. The saturation solubility of a solid is given by the fundamental thermodynamic relationships for equilibrium between a solid and a liquid phase. When the effect of differences between the heat capacities of the liquid and solid is neglected, the saturation solubility is given by: ln (γS xS) =

ΔHmelt 1

(T

R

−

1 Tmelt)

(8.2)

where xS is the solid solubility (mole fraction) in the solvent, γS is the liquid phase activity coefficient, ΔHmelt is the enthalpy of melting, Tmelt(K) is the solid melting temperature and T(K) is the system temperature. For ideal solutions, Eq. 8.1 reduces to the Vanʼt Hoff relationship: ln xS =

ΔHmelt 1 R

(T

−

1 Tmelt)

(8.3)

With this equation the solute solubility for an ideal solution can be calculated from the heat of fusion and the pure-solid melting temperature alone. In this case the solubility depends only on the properties of the solute and is independent of the nature of the solvent. A common type of solid-liquid phase diagram is the binary eutectic, shown in Fig. 8.5, in which a pure solid component is formed by cooling an unsaturated solution until solids appear. Continued cooling will increase the yield of pure component. At the eutectic point both components solidify and additional purification is not possible.

8.3 Crystallization from solutions

213

Fig. 8.5 Eutectic solid-liquid phase diagram.

Fig. 8.6 Solubility of some inorganic solutes in water.

Solution crystallizers are generally classified into one of five categories according to the method by which supersaturation is achieved. Fig. 8.6 illustrates the effect of temperature on the solubility of different inorganic solutes. The common techniques for producing solids from a solution include:

214

8 Crystallization and precipitation

• Lowering the temperature of the feed solution by direct or indirect cooling. If solute solubility is strongly temperature dependent (KNO3), this is the preferred approach • Adding heat to the system to remove solvent and thus “salt out” the solute. This technique is effective if solubility is insensitive to temperature (NaCl) • Vacuum cooling the feed solution without external heating. If solubility is strongly dependent on temperature this method is attractive • Combining techniques. Especially common is vacuum cooling supplemented by external heating for systems whose solubility has an intermediate dependence on temperature (Na2SO4) • Adding a non-solvent. This is a common technique for precipitating solute from solution and is useful as both a laboratory technique and as an industrial process for product recovery. These methods can be employed in single- or multistage crystallization or in batch operations. Generally, production rates over 50 ton/day justify continuous operation. The continuous operations tend to have higher yield and require less energy than batch, but batch is more versatile. Multistage operation is employed where evaporative requirements exceed the capabilities of a single vessel and/or energy costs dictate staging of the operation. Another reason for staging may sometimes be the production of more uniform and/or larger crystals. Operation of crystallizers in series generates crystal size distributions having narrower size spread than the same volume of crystallizers in parallel. Batch crystallizers produce a narrower CSD than continuous well-mixed units. Although numerous design methods and kinetic theories exist to analyze specific crystallizer configurations, no clear-cut guidance is available for the choice between crystallizer types and/or modes of operation. 8.3.2 Cooling crystallizers Cooling crystallizers obtain supersaturation by cooling the hot saturated solution. They are very desirable when they work because of their high yield and low energy consumption. Cooling crystallization is applicable to systems where temperature has a large influence on solubility (e.g. ammonium alum and soda). The method is not useful for systems like NaCl where temperature has little effect on solubility. When cooling is the selected mode by which supersaturation is generated, heat can be transferred through an external cooling surface or through coils or a jacket internal to the crystallizer body. The forced circulation crystallizer is a simple highly effective unit designed to provide high heat-transfer coefficients in either an evaporative or cooling mode. A schematic diagram in which slurry is withdrawn from the crystallizer body and pumped through an external heat exchanger is shown in Fig. 8.7. Most systems are operated with a high recirculation rate to provide good mixing inside the crystallizer and high rates of heat

8.3 Crystallization from solutions

215

Fig. 8.7 Schematic of forced circulation crystallizer with external cooling.

transfer to minimize encrustation. The tendency of the solute to form encrustations on the cooling surface often limits their operation by restricting the temperature of the cooling liquid and the temperature decrease of the slurry flowing through the heat exchanger. High heat transfer rates reduce the formation of encrustations considerably by minimizing the temperature difference over the heat transfer surface. It is not uncommon to limit the decrease in magma temperature to about 3 to 5 °C. The feed is commonly introduced into the circulation loop to provide rapid mixing with the magma (mixture of crystals and solution) and minimize the occurrence of regions of high supersaturation, which can lead to excessive nucleation. The use of a conventional heat exchanger can be avoided by employing directcontact cooling (DCC) in which the product liquor is allowed to come into contact with a cold heat-transfer medium. Other potential advantages include better heat transfer and smaller cooling load. However, problems include product contamination from the coolant and the cost of extra processing required recovering the coolant for further use. Crystallization processes employing DCC have been used successfully in recent year for dewaxing of lubricating oils, desalination of water and production of inorganic salts from aqueous solutions.

8.3.3 Evaporating and vacuum crystallizers Evaporative crystallizers supersaturate the solution by removing solvent through evaporation. These crystallizers are used when temperature has little effect on

216

8 Crystallization and precipitation

Fig. 8.8 Multiple-effect evaporating crystallizer cascade for ammonium sulfate.

solubility (e.g. NaCl) or with inverted solubilities (e.g. calcium acetate). The crystallizer may be as simple as a shallow open pan heated by an open fire. Steamheated evaporators are widely used to produce common salt from brine and sugar refining. These applications often use an evaporative crystallizer containing calandria (steam chest1, see also Fig. 2.11). The magma circulates by dropping through the central downcomer and then rises as it is heated in the calandria. At the top some of the solution evaporates increasing the supersaturation causing crystal growth. On large-scale also many types of forced circulation evaporating crystallizers are used. These systems are similar to the one shown in Fig. 8.7 and obviously very similar to evaporators without crystallization. Operational problems with evaporative crystallizers can be caused by scale formation on the heat exchanger surfaces or at the vapor-liquid interface in the crystallizer. Such problems can be overcome by not allowing vaporization or excessive temperatures within the exchanger and by proper introduction of the circulating magma into the crystallizer. The latter may be accomplished by introducing the magma below the surface of the magma in the crystallizer. The evaporators are often connected together to make a multiple-effect system as shown in Fig. 8.8. Here the vapor from the first unit is used as the steam to heat the second unit. This reduces the steam requirements per kg of product. The pressure must be varied so that the condensing vapor will be at a temperature greater than the boiling temperature in the second stage. A variety of ways connecting the different stages have been developed. Vacuum crystallizers utilize evaporation to both concentrate the solution and cool the mixture. They combine the operating principles of evaporative and cooling crystallizers. The hot liquid feed enters the crystallizer at a higher pressure and temperature, where it flashes and part of the feed evaporates. This flashing causes adiabatic cooling of the liquid. Crystals and mother liquor exit the bottom 1

Shell-and-tube unit used as a reboiler for evaporation (or distillation).

8.3 Crystallization from solutions

217

of the crystallizer. Because of the vacuum equipment required to operate at pressures of 5 to 20 mbar, the vacuum crystallizers are considerably more complex than other types of crystallizers and most common in large systems. 8.3.4 Continuous crystallizers Many different continuously operated crystallizers are available. The majority can be divided into three basic types: forced circulation, fluidized bed (OSLO) and draft-tube agitated units. A forced-circulation crystallizer has already been discussed and is shown in Fig. 8.7. Fig. 8.9 shows a draft-tube-baffle (DTB) crystallizer designed to provide preferential removal of both fines and classified product. A relatively low speed propeller agitator is located in a draft tube, which extends to a few inches below the liquor level in the crystallizer. The steady movement of magma up to the surface produces a gentle, uniform boiling action over the whole cross-sectional area of the crystallizer. Between the baffle and the outside wall of the crystallizer agitation effects are absent. This provides a settling zone that permits regulation of the magma density and control of the removal of excess nuclei. Flow through the annular zone can be adjusted to only remove crystals below a certain size and dissolve them in the fines dissolution exchanger. Feed is introduced to the fines circulation line to dissolve nuclei resulting from feed introduction.

Fig. 8.9 Schematic of a draft-tube-baffled (DTB) crystallizer.

218

8 Crystallization and precipitation

Fig. 8.10

Oslo crystallizer.

Another type of continuous crystallizer is the Oslo fluidized-bed crystallizer shown in Fig. 8.10. In units of this type a bed of crystals is suspended in the vessel by the upward flow of supersaturated liquor in the annular region surrounding a central downcomer. The objective is to form a supersaturated solution in the upper chamber and then relieve the supersaturation through growth in the lower chamber. The use of the downflow pipe in the crystallizer provides good mixing in the growth chamber. Crystal size distributions produced in a perfectly mixed continuous crystallizer are highly constrained. The form of the CSD in such systems is entirely determined by the residence time distribution of a perfectly mixed crystallizer. Greater flexibility can be obtained through introduction of selective removal devices that alter the residence time distribution of materials flowing from the crystallizer. Clear-liquor advance is simply the removal of mother liquor from the crystallizer without simultaneous removal of crystals. The primary objective of classifiedfines removal is preferential withdrawal of crystals whose size is below some specified value. A simple method for implementation of classified-fines removal is to remove slurry from a settling zone in the crystallizer. Constructing a baffle that separates the zone from the well-mixed region of the vessel can create the settling zone. The separation of crystals in the settling zone is based on the dependence of settling velocity on crystal size. Such crystals may be redissolved

8.3 Crystallization from solutions

219

and the resulting solution returned to the crystallizer. Classified-product removal is carried out to remove preferentially those crystals whose size is larger than some specified value. 8.3.5 Basic yield calculations For crystallization a high recovery of refined solute is generally the desired design objective. Similar to absorption (Chapter 3) and extraction (Chapter 5) it is common in crystallization to use solute free solvent flow rates and solute weight ratios in calculations. In this chapter we limit ourselves to the formation of pure crystals that contain no solvent. Mass balances concerning solvated (hydrated) crystals, though more complex, follow the same line of reasoning. The theoretical maximum product recovery or yield from the crystallizer is then defined by: Yield =

F ′ XF − S ′ XC F ′ XF

=

C F ′ XF

=1−

S ′ XC F ′ XF

(8.4)

where the meaning of the quantities F ′, XF , S ′ and XC are illustrated in Fig. 8.11 and defined as: XF = weight ratio of dissolved solute in the feed per unit mass of solvent (kg solute/kg solvent) XC = weight ratio of dissolved solute in solvent discharged from the crystallizer per unit mass of discharged solvent (kg solute/kg solvent) F ′ = solute free mass feed rate of solvent into crystallizer (kg solvent/hr) S ′ = solute free mass discharge rate of solvent leaving crystallizer (kg solvent/hr) The solute free solvent feed rate F ′ and the weight ratio of dissolved solute to solvent in the feed XF are usually fixed by upstream requirements. The weight ratio of dissolved solute in the liquid leaving the crystallizer XC is determined by solubility, which is set by the temperature of the operation. The other adjustable variable is the solute free solvent discharge rate S ′. In evaporative systems, adding heat to the system evaporates a fraction ϕE of the solvent. Thus S ′ decreases and recovery increases as more solvent is evaporated and removed as a separate product stream. The conclusion from the preceding discussion is that for a given feed concentration and solubility relationship the mode of crystallizer operation governs the maximum recovery of solute. For a cooling crystallization system, where no solvent is removed, the only control that affects maximum solute recovery is crystallizer temperature. For evaporative systems, recovery is influenced by both the system temperature and by the quantity of solvent vaporized and removed from the system. For a system wherein a pure solid component is crystallized by cooling or evaporation the overall mass balance provides the maximum solute yield and crystal production rate. From the overall material balance it follows that:

220

8 Crystallization and precipitation

Fig. 8.11

Mass balance for a binary crystallization system.

solute free solvent in + dissolved solute in = evaporated solvent out + solute free solvent out + dissolved solute out + crystals out

(8.5)

F ′ + F ′ XF = V ′ + S ′ + S ′ XC + C

(8.6)

where V ′, S ′ and C are the vapor rate, solute free solvent rate and crystal removal rate. The vapor rate and solute free solvent out are obtained from a mass balance over the solvent and become: V ′ = F ′ϕE

S ′ = F ′ (1 − ϕE)

(8.7)

in which ϕE is the ratio of the mass of solvent evaporated per unit mass of solvent fed. Introduction in Eq. 8.7 provides the maximum obtainable yield: Ymax = 1 −

XC XF

(1 − ϕE)

(8.8)

For specific operating modes this simplifies to: Cooling crystallization only

ϕE = 0 ⇒ Ymax = 1 −

Evaporation only

XC = XF ⇒ Ymax = ϕE

XC XF

(8.9)

(8.10)

8.4 Crystallizer modeling and design

221

Substitution of Eq. 8.7 into Eq. 8.6 and rewriting gives the crystal production rate C: C = F ′ (XF − XC (1 − ϕE))

(8.11)

The preceding expressions can be solved provided the composition of the exit stream is known. In many instances it is acceptable to assume that the exit stream composition corresponds to saturation conditions. Systems in which this occurs are said to exhibit fast growth or Class II behavior. Should growth kinetics be too slow to use all the supersaturation, the system is said to exhibit slow-growth behavior and is classified as a Class I system.

8.4 Crystallizer modeling and design When one is purely interested in the yield of a certain crystallization process, solving the mass balances for the solvent, solutes and the solids will be sufficient. Solubility and phase relationships influence the choice of crystallizer and method of operation. Although equilibrium and equilibrium calculations are very important, they alone are not sufficient for a complete crystallizer design as for other equilibrium staged separations. Equilibrium calculations can tell the total mass of crystals produced, but cannot tell the size and the number of crystals produced. However, as soon as the crystal size distribution (CSD) of the product becomes of interest, nucleation and crystal growth rates must be incorporated together with population balances. Nucleation leads to the formation of crystals. Growth is the enlargement of crystals caused by deposition of solid material on an existing surface. The relative rates at which nucleation and growth occur determine the crystal size distribution. Acceptable operating conditions for the minimization of uncontrolled nucleation and encrustation of heat-exchange surfaces are defined by metastable limits. 8.4.1 Supersaturation and metastability The kinetic phenomena that influence crystal size distributions are nucleation and growth. The driving force for both these phenomena is supersaturation, defined as the deviation from thermodynamic equilibrium. If a solution or melt are slowly cooled the temperature will gradually fall until the saturation point is reached. At this point however, no spontaneous solidification will take place because the rate of nucleation is negligible. Change of phase only occurs when a certain degree of supercooling provides the required driving force to initiate nucleation. This region of supercooling is known as the metastable region where the solution contains more dissolved solute than that given by the equilibrium saturation value.

222

8 Crystallization and precipitation

Fig. 8.12 Schematic depiction of supersaturation.

This is illustrated in Fig. 8.12 showing the solubility curve, the supersaturation line of a solution and the metastable region in which the nucleation rate is very low. The degree of supersaturation can be expressed in may different ways. In crystallization is common to use the difference between the solute concentration and the concentration at equilibrium: Δc = c − c ∗

(8.12)

where c is the actual solution concentration and c* the equilibrium saturation value. Another common expression is the relative supersaturation s, given by the ratio of the difference between the solute concentration and the equilibrium concentration to the equilibrium concentration: s=

c − c∗ c∗

(8.13)

Solution concentration may be expressed in a variety of units. For general mass balance calculations kilograms anhydrate per kilogram of solvent or kilograms hydrate per kilograms of free solvent are most convenient. The former avoids complications if different phases can crystallize over the temperature range considered. The importance of supersaturation in setting nucleation and growth rates is clarified in Fig. 8.13, which schematically shows the influence of supersaturation on nucleation and growth. The key aspects in this figure are the qualitative relationships of the two forms of nucleation to growth and to each other. Growth

8.4 Crystallizer modeling and design

Fig. 8.13

223

Influence of supersaturation on nucleation and crystal growth rates.

rate and secondary nucleation kinetics are low-order (shown as linear) functions of supersaturation, while primary nucleation follows a high order dependence on supersaturation. Design of a crystallizer to produce a desired crystal size distribution requires the quantification of nucleation and growth rates to externally controlled variables and to supersaturation. 8.4.2 Nucleation In crystallization, nucleation is the formation of a solid phase from a liquid phase. The process differs from growth in that a new crystal results from the transfer of solute from the liquid to the solid. Because it is the phenomenon of crystal formation, nucleation sets the character of the crystallization process, and it is therefore the most critical component in relating crystallizer design and operation to crystal size distribution. In primary nucleation existing crystals are not involved in the nucleation process. Homogeneous primary nucleation occurs in the bulk liquid phase in the absence of crystals or any foreign particles. Heterogeneous primary nucleation occurs on a solid surface such as a dust particle or the vessel wall. Secondary nucleation is heterogeneous nucleation induced by existing crystals. Homogeneous primary nucleation is only observed when the vessel is meticulously cleaned, polished and closed to avoid the presence of atmospheric dust that will act as heterogeneous nucleation sites. In classical homogeneous nucleation theory a large number of solute units (atoms, molecules or ions) can aggregate

224

8 Crystallization and precipitation

together in a highly supersaturated solution to form a cluster. For small enough particles the solute concentration may greatly exceed the normal equilibrium saturation value and the relationship between stable particle size and solubility takes the form of a Kelvin equation, similar to Eq. 7.17: ln

c(r)

[

c

∗

]

= ln (s + 1) =

Vm γ

⋅

2

(8.14)

ν⋅RT r

where c (r ) is the solubility of particles with radius r, c* the normal equilibrium solubility of the substance, s the supersaturation (see Eq. 8.13), γ the interfacial tension of the crystal surface in contact with its solution and Vm the molar volume. The quantity ν represents the number of moles of ions formed from one mole of electrolyte, for a nonelectrolyte ν = 1. Bringing together the solute units and expelling the solvent molecules requires work that serves as an energy barrier controlling the nucleation kinetics. Consideration of the energy involved in solidphase formation and in creation of the surface of an arbitrary spherical crystal of radius r in a supersaturated fluid leads to the relationship: ΔG = ΔGcrystal + ΔGsurface =

4π

(3 )

r 3ΔGL ⇒ S + 4 π r 2 γ

(8.15)

where ΔG is the overall excess free energy associated with the formation of the crystalline body and ΔGL⇒S is the free energy change associated with the phase change. If the cluster cannot reach a certain critical size rcr , ΔG is still positive and it redissolves. If the cluster can reach the critical size r = rcr , the free energy change upon further growth will be negative. The cluster is then stable and serves as a nucleus for further growth with nucleation rate B 0 = A · exp (–ΔGcr /kT). The critical size rcr is reached when d (ΔGcrystal +ΔGsurface)/dr = 0; combination with Eq. 8.15 gives rcr = –2 γ /ΔGL⇒S and, after elimination of ΔGL⇒S , the governing free energy ΔGcr = 4 π rcr2 γ / 3. The critical size of nuclei determines their solubility c (rcr) and the supersaturation s, see Eq. 8.14. At small supersaturations, where ln (s + 1) ≈ s, the nucleation rate becomes B 0 = A exp −

(

16 π Vm2 γ 3 NAv 3 ν 2 (RT)3 s 2 )

with frequency factor A ≈ 1030 nuclei/s · cm3

(8.16)

This is the simplest rate expression for homogeneous primary nucleation. The high-order dependence on supersaturation is especially important as a small variation in supersaturation may produce an enormous change in nucleation rate and can lead to production of excessive fines when primary nucleation mechanisms are important.

8.4 Crystallizer modeling and design

225

In heterogeneous primary nucleation the crystal forms around a foreign object such as a dust particle or a crack in the vessel wall. The foreign object lowers the surface energy because the solvent wets the solid. Although Eq. 8.16 is applicable, the frequency factor decreases by 4 to 5 orders or magnitude. This greatly increases the rate of nucleation at much lower supersaturations, which is of especial importance to industrial systems that are never completely free of suspended solids, less perfectly cleaned, and contain incompletely polished interior surfaces. Secondary nucleation is the formation of new crystals as a result of the presence of solute crystals through several mechanisms including initial breeding, contact nucleation, and sheer breeding. In most industrial crystallizers secondary nucleation is far more important than primary nucleation. This is because continuous and seeded batch crystallizers always contain crystals that can participate in secondary nucleation mechanisms and are operated at low supersaturation to obtain regular and pure product. The presence of crystals and low supersaturation can only support secondary nucleation and not primary nucleation. Initial breeding results from adding seed crystals to a supersaturated solution to induce nucleation. The presence of growing crystals can also contribute to secondary nucleation by several different mechanisms. Contact nucleation results from nuclei formed by collisions of crystals against each other, against crystallizer internals or with an impellor, agitator or circulation pump. Sheer breeding results when a supersaturated solution flows by a crystal surface and carries along crystal precursors formed in the region of the growing crystal surface. Many commercial crystallizers operate in the secondary nucleation range. The ease with which nuclei can be produced by contact nucleation is a clear indication that this mechanism is dominant in many industrial crystallization operations. Based on experimental observations the contact nucleation rate was proposed to follow an exponential function of impact energy E minus a required minimum collision energy Et to break small nuclei from the growing crystals: B 0 = kN exp (E − Et)

E > Et

(8.17)

Although this equation can be derived from fundamental principles, it is only accurate within an order of magnitude. For design purposes the metastable limit can provide a useful semi-empirical approach to correlate the effective or apparent heterogeneous and secondary nucleation rate: n

n

B 0 = kn (c − cm) ≈ kn (c − c ∗)

c ∗ < cm

(8.18)

where c is the solute concentration, cm is the solute concentration at which spontaneous nucleation occurs and c ∗ is the solute concentration at saturation. The metastable limit must be determined through experimentation. Fortunately cm is very close to c ∗ for many inorganic systems. The parameters kn and n must be evaluated from experimental data. The order for heterogeneous nucleation can range from 2 to 9. For secondary nucleation the order is significantly lower and

226

8 Crystallization and precipitation

in the range of 0 to 3. Much of the experimental data is correlated by an empirical power-law function: B 0 = kn′ MTn G m

(8.19)

where MT = magma concentration [crystal weight per unit suspension volume] and G = crystal growth rate [m per unit time]. The orders m and n can be determined from small scale experiments; a reliable value of kn′ , however, should be determined from data for commercial crystallizers. 8.4.3 Crystal growth Crystal growth is a complex subject, which, like nucleation, is not completely understood and may be expressed in a variety of ways: • linear advance rate of an individual crystal face • change in a characteristic dimension of a crystal • rate of change in mass of a crystal. Adsorption layer or kinetic theories have proven to be quite fruitful in explaining crystal growth. They postulate that there is an “adsorbed” layer of units, which is loosely held to the crystal face. These adsorbed units are free to move on the twodimensional surface, but they have essentially lost one degree of freedom. Thus the unit must crystallize through two-dimensional nucleation. At low supersaturations the energy required for two-dimensional nucleation is considerably less than for normal nucleation. Thus, existing crystals can grow under conditions where three-dimensional nucleation will not occur. Crystal growth will usually occur at supersaturation levels lower than can be explained by the two-dimensional nucleation theory. This happens because growth is much faster at any imperfection where the surface is not a perfect plane. Since crystals are usually not perfect, growth usually proceeds by a “filling-in” process. Small crystals are more likely to be perfect than large crystals and therefore more likely to require two-dimensional nucleation. Large crystals are less likely to be perfect, and are more likely to be damaged by the impeller or baffles. Thus, the large crystals can grow by healing kinks, pits and dislocations. These imperfections are also the reason that even two crystals of the same size may have different growth rates although all conditions appear to be the same. This is called growth rate dispersion. If one of the crystals happens to be more perfect than the other it will have a lower growth rate. The adsorption theory that explains the crystallization once a unit is at the surface is often incorporated into mass transfer theories that describe the movement of the unit to the surface. In crystallization the picture of the mass transfer process is somewhat different from mass transfer theories for other processes and

8.4 Crystallizer modeling and design

Fig. 8.14

227

Schematic of crystal growth mass transfer process.

different empirical expressions are used. Fig. 8.14 shows a schematic of the mass transfer process that can be broken down in the following seven steps: 1. 2. 3. 4. 5. 6. 7.

Bulk diffusion of solvated units through the film Diffusion of solvated units in the adsorbed layer Partial or total desolvation Surface diffusion of units to growth site Incorporation into lattice Counter diffusion of solvent through the adsorbed layer Counter diffusion of solvent through the film

Since it is difficult to include all these steps in a theory, usually a simplified model is used which includes step 1, combines step 2 to 5 and ignores the last two steps. The mass transfer across the film is for step 1 is given by: dm dt

= k f A (c − ci)

(8.20)

where kf is the film mass transfer coefficient and A the crystal surface area. Because m represents the mass of solute transferred across the film, m also equals the mass of solid deposited. Next, steps 2 to 5 are combined as a single surface “reaction” of “order” n with rate constant kr through the empirical relation: dm dt

= k r A(ci − c ∗) n

(8.21)

In general it is very difficult or impossible to determine the interfacial concentration ci . For a first order surface reaction (n = 1) both equations can be combined to remove ci in a similar way as shown in chapter 4 for gas/liquid contactors:

228

8 Crystallization and precipitation

dm dt

= kOG A(c − c ∗)

(8.22)

in which kOG is the overall rate constant. For crystals where the faces are growing at the same rate, the mass and area of the crystal can be written as m = k v L 3 ρc

A = kA L 2

(8.23)

where kv and kA are shape factors. Substituting into Eq. 8.12 we obtain the linear growth rate G of the crystal, which is equal in all three dimensions: G=

dL dt

=

kA

(3 ρc kV)

kOG (c − c ∗)

(8.24)

that can be simplified by the introduction of the combined growth rate constant kG to: dL = kG (c − c ∗) (8.25) G= dt When the linear growth rate is independent of crystal size and the supersaturation is constant, this is reduced to the very simple ΔL-law, which will be used in the next section: ΔL (8.26) G= Δt Because of the assumptions involved, it should not be surprising that many systems do not satisfy this law. In addition to supersaturation temperature will also affect the crystal growth rate through diffusion and the crystal rearrangement process. If either mass transfer or surface reaction controls the temperature dependence of growth kinetics can often be expressed in terms of an Arrhenius expression: kG = kG0 exp −

ΔEG

( RT)

(8.27)

When both mechanisms are important this equation is not valid and Arrhenius plots for crystal growth data tend to give curved instead of straight lines. When n is not a simple integer, it is not possible to solve explicitly for the growth rate. In these cases a common approach is to introduce an empirical “overall growthrate order” n to relate growth kinetics to supersaturation: dm dt

= kOG A (c − c ∗) n

(8.28)

yielding for the growth rate a power-law function in which the constants kG and n are determined by fitting experimental crystal growth data:

8.4 Crystallizer modeling and design

G=

dL dt

= kG (c − c ∗) n

229

(8.29)

Such an approach will only be valid over small ranges of superstaturation. 8.4.4 Population balance equations for the MSMPR crystallizer An analysis of a crystallization system requires the quantification of nucleation and growth kinetics, mass and energy balances and the crystal size distribution. There are a number of ways of describing the crystal size distribution for a crystallizer. Population balances are the major theoretical tool for predicting and analyzing crystal size distributions. They allow one to quantify the nucleation and growth processes by balancing the number of crystals within a given size range. A balance on the number of crystals in any size range, say L1 to L 2, accounts for crystals that enter and leave the size range by convective flow into and out of the magma suspension with volume VM and for crystals that enter and leave the size range by growth. Crystal breakage and agglomeration are ignored and it is assumed that crystals formed by nucleation are near to size zero. The number of crystals in the size range L1 to L 2 is given by: L2

∫

ΔN = n d L

(8.30)

L1

The population balance differs from a mass balance in that only the item balanced within a given size range is included, which are coupled to the more familiar balances on mass and energy. It is assumed that the population distribution is a continuous function and that a characteristic dimension L can describe crystal size, surface area and volume. Population balances for crystallizers are usually described by the population density n that gives the number of crystals dN in the size range L to L + dL by: n=

dN dL

= lim ΔL → 0

ΔN ΔL

(8.31)

where ΔN is the number of crystals in size range ΔL per unit volume (clear liquor or slurry). The value of N depends on the value of L at which the interval ΔL is measured as shown in Fig. 8.15. In practice, the number density is calculated from the relationship: n=

mi kv ρc L¯¯¯i3ΔL i

(8.32)

230

8 Crystallization and precipitation

Fig. 8.15 Schematic representation of the population density.

Fig. 8.16 Schematic diagram of a simple, perfectly mixed crystallizer.

where mi is the mass of the residue on sieve i, ¯ L i is the mean crystal size of the particles in that fraction of the sieve and ΔLi is the size difference between the two sieves that produced mass fraction mi . Application of the population balance is most conveniently described for the idealized case of a continuous, steady state MSMPR (mixed-suspension mixedproduct removal) crystallizer shown in Fig. 8.16. Furthermore many industrial crystallizers are operated in a well-mixed manner and the equations derived are quite useful in describing their performance. Perfectly mixed crystallizers are highly constrained and the form of crystal size distribution produced by such systems is fixed. The assumptions made are that no crystals are present in the

8.4 Crystallizer modeling and design

231

feed stream, all crystals within a size class are of the same size and crystal growth rate is independent of crystal size. For a well-mixed and constant slurry volume VM in a crystallizer the population balance for a time interval Δt and size range ΔL = L2 – L1 becomes: number of number of number of crystals growing = crystals growing + crystals in range (8.33) into range out of range leaving crystallizer which becomes in symbols: nL G VM Δt

= nL + ΔL G VM Δt

+ Qn¯ ΔL Δt

(8.34)

where Q is the volumetric feed and discharge rate, G is the growth rate, defined in Eq. 8.26, and n¯ is the average population density. dn ΔL , the population balance As ΔL → 0, n¯ → n, and, noting that nL + ΔL = nL + dL takes the form G

dn dL

Qn

=−

(8.35)

VM

which can be written as: G

dn dL

+

n

=0

τ

(8.36)

by introducing a mean residence time τ defined as VM /Q. This equation can be integrated using the boundary condition n = n0 at L = 0 to obtain: −

n = n0 e (

L Gτ)

(8.37)

If the logarithm of the number density n is plotted against the crystal size L, a straight line is produced with the negative slope – (1/Gτ) and intercept ln n0 . Under steady-state conditions the total number production rate of crystals in a perfectly mixed crystallizer must be identical to the nucleation rate B 0. With the intercept of the ordinate n0 for L = 0, we obtain the following for the nucleation rate B 0: B0 =

dN

( dt )

L=0

=

dN

dL

( dL ) ( dt ) L=0

∞

2

(8.38)2

= n0 G

Note that the total number of nuclei per unit volume N =

∫ 0

∞

∫

n ⋅ dL = n0 e 0

−L / Gτ

0

dL = n0 Gτ = B τ.

232

8 Crystallization and precipitation

The slope –(1/Gτ) and the intercept of the ordinate n 0 of the straight line in the number-density diagram can thus be used to determine the two kinetic parameters: nucleation rate B 0 and growth rate G. These two parameters determine the dominant crystal size LD of the crystal size distribution. The dominant crystal size LD is defined as the crystal size for which the maximum weight in the crystal size distribution is obtained. The crystal mass mL within the range of L to L + dL is obtained from the mass mc of a single crystal multiplied by the number of crystals N (Eq. 8.30) within that population: −

mL = (ρc kv L 3) N = ρc kv n0 L 3 e(

L Gτ)

(8.39)

The maximum weight is then obtained by setting the derivative of mL to L, dmL , equal to zero and solving this equation for L. The resulting dominant ( dL ) crystal size LD is then: LD = 3 Gτ

(8.40)

In a similar way the suspension density of total crystal (magma) mass MT (kg crystals/m3 suspension) is calculated from the mass of one crystal mc with a size between L and L + dL, (Eq. 8.23) and integrating this over the total number of crystals NT (Eqs. 8.30 en 8.37): ∞

∫

−

MT = ρc kv n 0 L 3 e(

L Gτ)

dL = 6 ρc kv n 0 (Gτ)4

(8.41)3

0

which is easily converted into the volumetric holdup of crystals ϕT (m3 crystals / m3 suspension) by dividing with the crystal density: ϕT = 6 kv n0 (Gτ)4

(8.42)

8.5 Other crystallization techniques 8.5.1 Precipitation Precipitation is an important industrial process, which is closely related to crystallization from solution. In precipitation a solute is made supersaturated so that dissolved solutes precipitate out. Precipitation differs from crystallization since ∞

3

Note that

∫ 0

L

3

e

L − ( Gτ)

dL = (Gτ)

∞ 4

∫y 0

3

e

−y

4

dy = 6 (Gτ) .

8.5 Other crystallization techniques

233

the precipitate is usually amorphous, has a poorly defined size and shape, is often an aggregate, and is usually not pure. Precipitation can be carried out in one step to remove a large number of compounds in a single precipitate. This is the easiest application and is most common. Fractional precipitation requires a series of steps with each step optimized to precipitate the desired component. It is not possible to do extremely sharp purifications by fractional precipitation. Precipitation and crystallization are often complementary processes since precipitation is used for a “rough cut” while crystallization is used for final purification. Precipitation is commonly used for the manufacture of pigments, pharmaceuticals and photographic chemicals. In the production of ultra fine crystalline powders, precipitation is often considered an attractive alternative to comminution, particularly for heat sensitive substances. Like all crystallization processes, precipitation consists of three basic steps. The creation of supersaturation, followed by the generation of nuclei and the subsequent growth of these nuclei to visible size. The used equipment is often very similar to that used for crystallization. Precipitation operations usually have a mixing operation where various reagents are added to make the solution supersaturated. This is followed by a holding period, which allows for an induction period before nuclei form and a latent period before the supersaturation starts to decrease. Once the precipitate starts to grow the desupersaturation often occurs at a constant rate. Towards the end of the desupersaturation period the rate decreases as the concentration approaches equilibrium. During this period aging often occurs. During aging small crystals redissolve and the solute is redeposited onto the larger precipitates. The result is a uniform crop of fairly large precipitate, which is relatively easy to separate from the solution by centrifugation. Precipitation processes are conveniently classified by the method used to produce the supersaturation: cooling, solvent evaporation, reaction, salting out, antisolvents and pH. Supersaturation caused by cooling and or evaporation of solvent is very similar to solution crystallization. Many products are made by mixing reagents that react to form an insoluble precipitate. A representative example is the formation of gypsum by the addition of sulfuric acid to a calcium chloride solution: CaCl2 + H2SO4 → CaSO4 ↓ + 2 HCl In aqueous solution high inorganic salt concentrations also often cause the precipitation of solute. Antisolvents or non-solvents can be added to achieve that same effect. The solubility of many compounds is effected by pH. 8.5.2 Melt crystallization In some crystallization systems the use of a solvent can be avoided. Melt crystallization is the process where crystalline material is separated from a melt of the crystallizing species by direct or indirect cooling until crystals are formed in the

234

8 Crystallization and precipitation

liquid phase. Two basic techniques of melt crystallization are gradual deposition of a crystalline layer on a chilled surface or fast crystallization of a discrete crystal suspension in an agitated vessel. Zone melting relies on the distribution of solute between the liquid and solid phases to effect a separation. Normal freezing is the slow solidification of a melt. The impurity is rejected into the liquid phase by the advancing solid interface. The basic requirement of melt crystallization is that the composition of the crystallized solid differs from that of the liquid mixture from which it is deposited. The process is usually operated near the pure-component freezing temperature. High or ultrahigh product purity is desired in many of the melt purification processes. Single-stage crystallization is often not sufficient to achieve the required purity of the final product. Further separation can be achieved by repeating the crystallization step or by countercurrent contacting of the crystals with a relatively pure liquid stream. Since there is no solvent, melt crystallization has the advantage that no solvent removal and recovery is required and contamination by the solvent is impossible. However, there is also no way to influence the melt properties (viscosity, diffusivity) and the chemicals being purified must be stable at the melting point. For salts with very high melting points solution crystallization is less expensive. At the lower temperatures possible in solution crystallization distribution coefficients can be orders of magnitude more favorable. These comparisons show why crystallization from solution is much more common for bulk separations. Melt crystallization becomes advantageous when the presence of a solvent would be detrimental or when highly pure products are desired. Currently the interest in wider application of melt crystallization is stimulated by the energy-saving potential in large scale processing because solvent evaporation is absent and the heat of fusion is several times less than the heat of vaporization. Multiple countercurrent contacting stages are possible by conducting melt crystallization inside a column in a manner somewhat analogous to distillation. The main incentive is to attain higher purity product than can be achieved in a single stage of conventional crystallization. The concept of a column crystallizer is to form a crystal suspension and to force the solids to flow counter currently against a stream of enriched reflux liquid by gravity or rotating blades. At the end of the crystallizer the crystals are melted. A portion of the melt is removed as product and the remainder is returned to the system as enriched reflux to wash the product crystals. One of the early column crystallizers, shown schematically in Fig. 8.17, was developed for the separation of xylene isomers. In this unit p-xylene crystals are formed in a scraped-surface chiller above the column and fed to the column. The crystals move downward counter currently to impure liquid in the upper portion of the column and melted p-xylene in the lower part of the column. Impure liquor is withdrawn from an appropriate point near the top of the column of crystals while pure product, p-xylene, is removed from the bottom of the column. An inherent limitation of column crystallization is the difficulty of control-

8.5 Other crystallization techniques

Fig. 8.17 Vertical column melt crystallizer.

Fig. 8.18

Sulzer MWB melt crystallizer system.

235

236

8 Crystallization and precipitation

ling the solid-phase movement, because the similar phase densities make gravitational separation difficult. The Sulzer MWB4 system, schematically depicted in Fig. 8.18, is an example of a commercial melt crystallization process that uses the gradual deposition of solids on a chilled surface. Crystal growth is on the inside of a battery of tubes through which melt is flowing. During crystallization the front of crystals advances into the direction of the mother liquor. This built up of a solids layer requires sequential operation. Steps include partial freezing of a falling film of melt inside vertical tubes, followed by slight heating, a “sweating” operation, and complete melting and recovery of the refined product. The recovered product melt can be put through the cycle again to increase purity or fresh feed can be introduced to the cycle. The process has been used on a large scale in the purification of a wide range of organic substances.

Nomenclature B0 cv c* C F′ G ΔEG ΔG ΔHmelt LD m MT

homogeneous primary nucleation rate (Eq. 8.16) coefficient of distribution equilibrium saturation concentration crystal production rate mass feed rate, solute free crystal growth rate energy of crystal growth excess free energy enthalpy of melting (dominant) crystal size population density in terms of mass total crystal mass (magma) (Eq. 8.41)

[nuclei / unit time · unit volume]

[–] [mol m–3] [kg / unit time] [kg / unit time] [m / unit time] [J mol–1] [J mol–1] [J mol–1] [m] [kg / unit volume] [kg crystals / unit volume suspension] or total crystal mass (magma) (Exercise 8.10) [kg crystals / unit volume solvent] n population density, Eq. 8.31 [number of nuclei m–1 m–3] n0 population density of nuclei (L = 0) [number of nuclei m–1 m–3] N nuclei concentration, Eq. 8.30 [number of nuclei / unit volume] Q volumetric discharge rate [m3 / unit time] r radius [m] s relative supersaturation [–] S′ mass discharge rate of solvent, solute free [kg / unit time] T (Tmelt) temperature (at melting point) [K] V′ vapor rate [kg / unit time] molar volume [m3 mol–1] Vm volume of magma (suspension or clear liquor) [m3] VM solubility of solids, mole fraction [–] xS X weight fraction [–] Y yield [–] 4

Metallwerk Bruchs

Exercises

γ γS ν σ τ ϕE ϕT

interfacial tension activity coefficient of solvent number of moles ions in one mole of electrolyte standard deviation of distribution residence time in crystallizer unit solvent evaporated per unit solvent fed volumetric holdup of crystals

237

[N m–1] [–] [–] [m] [unit time] [–] [unit volume crystals / unit volume suspension]

Exercises 1. Determine the maximum yield of anhydrous sugar crystals deposited at equilibrium in the following situations. Initially 100 kg of water is present. a. A saturated sugar solution is cooled from 80 °C to 20 °C. b. A saturated sugar solution at 80 °C has half of its water evaporated and the solution is then cooled to 20 °C Temperature: Solubility:

20 °C 2.04

80 °C 3.62 (kg anhydrous sugar / kg water)

2. A solution of 80 wt% of naphthalene in benzene is cooled to 30 °C to precipitate naphthalene crystals. The solubility of naphthalene at this temperature is 45 wt% naphthalene in solution. The feed rate is 5000 kg solution / h. a. Calculate the maximum production rate of crystals b. Calculate the maximum yield in kg crystals per unit weight of solute fed 3. A cooling crystallizer is used to crystallize sodium acetate, NaC2H3O2 (MW = 0.082 kg/mol), from a saturated aqueous solution. The stable form of the crystals is a hydrate with 3 moles of water. The feed is initially saturated at 40 °C while the crystallizer is cooled to 0°C until equilibrium conditions are established. If we initially dissolved the anhydrous salt in 100 kg of water/h, how many kg of crystals/h are collected? Temperature: Solubility:

0 10 20 30 40 60 (°C) 0.363 0.408 0.465 0.545 0.655 1.39 (kg anhydrous salt / kg water)

4. Copper sulfate is crystallized as CuSO4 · 5 H2O by combined evaporative / cooling crystallization. 1000 kg/h of water is mixed with 280 kg/h of anhydrous copper sulfate (MW = 0.160 kg/mol) at 40 °C. The solution is cooled to 10 °C and 38 kg/h of water is evaporated in the process. How many kg/h of crystals can be collected theoretically? Temperature: Solubility:

10°C 17.4

40°C 28.5 (g anhydrous salt / 100 g water)

5. Calculate the maximum (theoretical) yield of pure crystals that could be obtained from a saturated solution of sodium sulfate with 0.25 kg anhydrous salt / kg free water by a. isothermal evaporation of 25 % of the free water b. isothermal evaporation of 25 % of the original water c. cooling to 10 °C and assuming 2 % of the original water lost by evaporation; XC = 0.090 kg anhydrous salt / kg solute-free solvent

238

8 Crystallization and precipitation

d. adding 0.75 kg ethanol / kg free water; XC = 0.030 kg anhydrous salt / kg solutefree solvent e. adding 0.75 kg ethanol / kg free water; XC = 0.068kg Na2SO4 · 10 aq / kg solutefree solvent MW (Na2SO4 · 0 aq) = 0.142 kg/mol 6. Assuming that G∝Δc, determine whether or not the Arrhenius equation is valid for (NH4)2SO4 crystallization. If it is valid, determine ΔE. Note: c* depends on T and ammonium sulfate crystallizes as an anhydrate Given Temperature: Solubility: s: G:

30 60 90 °C 78 88 99 g anhydrous salt / 100 g water 0,05 0,05 0,01 – 5,0 8,0 0,6 10–7 m/s

7. Estimate the supersaturation s of an aqueous solution of K2SO4 at 30 °C required to get a heterogeneous nucleation rate of 1 nucleus per second per unit volume. Eq. 8.16 is applicable with an apparent interfacial surface tension γ of 2 · 10–3 Pa.s and a preexponential coefficient A = 1025 nuclei per second per unit volume. The density of the K2SO4 crystals is 2662 kg/m3, its molecular weight 0.174 kg/mol. 8. Potash alum (potassium aluminum disulfate) was crystallized from its aqueous solution in a laboratory scale continuous MSMPR crystallizer of 10 l capacity at steady state, supersaturation being achieved by cooling. The size analysis of the produced crystals resulting from a steady-state sample taken from the crystallizer operated at means residence time of 900 s, is given in the table below. Assume that the crystal density is 1770 kg/m3 and the volume shape factor is 0.47. Sieve Analysis Data:

Standard sieve Size (μm)

Weight on the sieve (g / kg of water)

850 710 500 355 250 180 125 90 63 45 Πosm (c)

(11.2)

A complication may be the increasing salt concentration in the remainder of the solution, causing the osmotic pressure to increase to such an extent that the re1 The Vanʼt Hoff relation πosm = c R T applies only to dilute solutions; c is the total amount of solute per unit pure solution, see Table 11.1.

11.4 Flux equations and selectivity

307

quired value of Pextern gets too high for practical purposes. As an example, Table 11.1 shows the osmotic pressure of various aqueous solutions. The production of potable water, which contains less than 1000 ppm of total dissolved solids2, requires lots of effort. Reverse osmosis plays an important role in providing an alternative to meet such a requirement, especially in places where conventional processes are not feasible. Small RO units find applications in remote areas like ships or islands, but RO-plants can be very large, producing up to 1 m3 s–1 of fresh water.

11.4 Flux equations and selectivity The performance of a membrane is determined by experimental conditions (such as pressure difference across a membrane), the geometry of a membrane module and, last but not least, by membrane properties. In this section, we focus on two important membrane properties: • the flux, how much of the component to be removed is passing through the membrane per unit time and per unit membrane area, and • the selectivity3, a measure of the relative permeation rates of different components through the membrane.

11.4.1 Flux definitions In this section, both properties will be addressed. To start with the former one, the flux Ji of a component i through a membrane is related to the driving force for transport. In case of a pressure difference across a membrane, the flux of component i, Ji , is defined through the following flux equation: Ji ≡ PM, i

p0 − pδ δM

= PM, i

Δpextern δM

(11.3a)

The flux, expressed in amount per unit time per unit surface area, is proportional to the absolute value of the – positive – pressure difference p0 – pδ over the membrane with thickness δM . The proportionality constant is called the permeability PM,i of component i through the chosen membrane, which is a unique value for a particular combination of membrane and component. In case the osmotic pressure cannot be neglected, the effective pressure difference for transport is decreased with the difference in osmotic pressures between feed and permeate solutions: 2 3

As defined by the World Health Organization. Also expressed is the fraction of solute in the feed retained by the membrane, known as the retention.

308

11 Membrane filtration

Ji = PM, i

Δpextern − ΔΠosm δM

(11.3b)

where ΔΠosm = Π0 – Πδ. In absence of a pressure difference across a membrane, the driving force for transport through the membrane may be the concentration difference. Then the related flux equation becomes Ji ≡ PM, i

c0 − cδ δM

(11.3c)

Permeability should be expressed in consistent units. Comparing the permeability with the coefficient of diffusion in Fickʼs equation and realizing that the dimensions of pressure p and concentration c differ by a factor R T, one may conclude that the dimension of PM,i is m2 s–1 divided by J mol–1 (the dimension of R T). However, for historical reasons and practical purposes, permeability, especially for gas separations, usually is expressed in barrer, defined as 1 barrer ≡

10−10 cm3 (STP) cm cm2 s cm H g

(11.4)

In other words, the permeability is a transport flux through a membrane per unit transmembrane driving force per unit membrane thickness. The factor 10–10 is added just for convenience. IUPAC applies the same definition, but recommends expressing permeability in SI units (kmol m m–2 s–1 kPa–1.) rather than in cgs units. An alternative way to express how fast a certain species can flow through a membrane is the permeance, which is defined as the ratio between permeability and membrane thickness or as a transport flux per unit transmembrane driving force. Thus permeance is a proportionality constant just like a mass transfer coefficient. If the driving force in the flux equation is expressed as concentration difference (see Eq. 11.3c), permeance is given in m s–1. The values of permeability are closely connected to molecular and/or membrane properties, depending on the type of mass transport through a membrane, which can be porous or dense. In porous membranes with relatively wide pores, viscous flow takes place with a pressure difference as driving force. In small pores (small compared to the mean free path) Knudsen flow is predominant. In some cases the diameter of a molecule exceeds that of the pore, resulting in so-called size exclusion. In dense membranes a solution-diffusion mechanism takes place. The different mechanisms are depicted in Fig. 11.6. Viscous flow does not offer any selectivity, whereas size exclusion only depends on molecular size relative to pore diameter. Therefore, in the following sections, relations for permeability and selectivity will be derived just for diffusion through porous membranes and for solution-diffusion through dense membranes.

11.4 Flux equations and selectivity

309

Fig. 11.6 Transport mechanisms in membranes. (a) viscous flow through macro- and mesopores; (b) diffusion through micropores; (c) size exclusion; (d) solution through a dense membrane.

11.4.2 Permeability for diffusion in porous membranes Mass transport in pores not only occurs by viscous flow. Molecular diffusion will contribute when a difference in concentration across the membrane exists. The flux equation for combined convection and diffusion in a binary mixture with components i and j is Ji = − ctot Dim

dxi dz

+

ci ctot

(Ji

+ Jj)

(11.5)

where Dim is the coefficient of diffusion of species i in the mixture. In case of equimolar counter diffusion the second term of the right hand side becomes zero and Equation 11.5 transfers into the well-known Fickʼs law: Ji = − ctot Dim

dxi dz

≈ Dim

ci,0 − ci, δ δM

(11.6)

This equation only applies to diffusional transport through membranes if the pressure on either side of the membrane is kept equal. However, Eq. 11.6 is still useful as an approximation for unimolecular diffusion in dilute systems (ci / ctot ≪ 1). For ideal gas mixtures Eq. 11.6 transfers into Ji = −

Dim dpi R T dz

≈

Dim pi, 0 − pi, δ RT

δM

(11.7)

By analogy with Eq. 11.3a, the permeability appears to be proportional with molecular diffusivity and the selectivity for diffusive transport through membranes is determined by the ratio of the diffusivities.

310

11 Membrane filtration

For ideal gases the diffusivity can easily be related to molecular properties by simple kinetic theory: Di =

1 3

λi ¯ vi

(11.8)

RT

where λi =

, the mean free path and ¯ vi =

√2 σi ptot NAv

8RT

√ π Mi

, the mean speed.

σi is cross-sectional area of species i and Nav the Avogadro number. Thus diffusivity is proportional to the reciprocal of the square root of molecular mass4 and the selectivity Sij is given by Sij =

Dim Djm

=

Mj

√Mi

(11.9)

When the pore diameter is of the same magnitude or smaller as the mean free path Knudsen diffusion is the main transport mechanism. In this case, collisions of molecules with the pore wall are more frequent than intermolecular collisions. Equation 11.7 is still valid as long as the Knudsen diffusivity DiK replaces the molecular diffusivity Dim . In straight cylindrical pores of diameter dp the Knudsen diffusivity is calculated from DiK =

dp 3

¯ vi

(11.10)

Unlike the value of Dim , the value of DiK is not just a molecular property, but also depends on the pore diameter dp . The selectivity, however, still obeys Eq. 11.9. In case of size exclusion, the larger molecule cannot permeate at all and the selectivity goes to infinity. 11.4.3 Permeability for solution-diffusion in dense membranes The transport mechanism in dense membranes is quite different from that determining transport through (micro)porous membranes and, actually, cannot be considered a special kind of filtration. Nevertheless, dense membranes are often applied and a basic knowledge is valuable. Dense membranes let permeating species dissolve in its matrix, creating a concentration gradient across the membrane thickness. The concentration gradient offers the driving force for dissolved molecules to travel to the other side of the membrane by molecular diffusion. Once at the other side, the dissolved molecules are freed and transferred into the permeate phase. Thus, permeability in this case will depend on both solubility and diffusivity, as will be quantified in the following. 4

Assuming that the ratio of cross-sectional areas is approximately unity.

11.4 Flux equations and selectivity

Fig. 11.7

311

Solution-diffusion mechanism in a dense membrane.

Unlike porous membranes, where no jump in concentration at the membrane surface exists, dense membranes show different concentrations at both sides of the membrane surface. Mass transport through the membrane takes place, the driving force being the concentration difference ci,0 – ci,δ . Both concentrations are defined at either membrane surface just inside the membrane. In Fig. 11.7 two possible concentration gradients are drawn. The usual situation is given by the solid line. Upstream, at the feed side, the feed concentration in the boundary layer just outside the membrane is assumed to be in equilibrium with the concentration of solute i, ci,0 , at the same surface just inside the membrane, see Fig. 11.7. The same assumption applies to the product side: the product concentration in the boundary layer just outside the membrane is supposed to be in equilibrium with the solute concentration just inside the downstream membrane surface. In the special case that no mass transfer limitation exists outside the membrane, the boundary values just outside the membrane equal the bulk feed concentration, ci,F , respectively the bulk product concentration, ci,P . Thermodynamics control the concentration ratio at both surfaces: Ki =

ci, 0 ci, F

=

ci, δ

(11.11)

ci, P

Elimination of the concentration difference inside the membrane from Eqs. 11.6 and 11.11 and, again, assuming ideal gas behaviour, leads to the following flux equation for dense membranes: Ji = Ki Dim

ci, F − ci, P δM

=

Ki Dim pi, F − pi, P RT

δM

(11.12)

Comparison with the original equation, Eq. 11.3a, shows that now the permeability in a dense membrane is proportional to the product of thermodynamic equilibrium partition constant, Ki , and the diffusivity through that membrane,

312

11 Membrane filtration

Dim . Eq. 11.12 is still valid in the usual case that mass transfer limitations outside the membrane do have an influence on the overall transport rate, as long as the upstream and downstream boundary concentrations are applied instead of the bulk values. Comparison of Eqs. 11.9 and 11.12 learns that the selectivity Sij found in separations with dense membranes will be much higher than those obtained with porous membranes: Sij =

Ki Dim Kj Djm

=

Ki

Mj

Kj

√Mi

(11.13)

The reason is that the ratio of solubilities, Ki /Kj , usually overrules the importance of molecular weights. A limiting situation arises when a certain species in a mixture cannot dissolve in the membrane material at all: the related selectivity goes up to infinity. 11.4.4. Selectivity and retention In membrane filtration, the selectivity is usually expressed as the retention of a certain species, by the formula: Ri =

CRi − CPi CRi

(11.14)

where Ri is the retention of species i, CRi the concentration of i in the retentate and CPi the concentration of i in the permeate. It is obvious that, if CRi equals CPi , there is no separation and Ri = 0. However, if CPi = 0, the retention of i is complete. Membranes are characterised by a Molecular Weight Cut-Off (MWCO) of, for instance, 60,000 dalton (Da). The definition of an MWCO of 60,000 Da is that the membrane retains 90 % of a compound with an MW of 60,000 Da. In Figure 11.8, the MWCO of two membranes are shown. Membrane A has a very shallow profile and Membrane B a sharp one. With Membrane B, insulin can easily be separated from human serum albumin, because the retention of insulin is 0 %, while the retention of human serum albumin is about 96 %. With Membrane A, the insulin retention is 32 %. This means that only 68 % of the insulin ends up in the permeate. With Membrane A, the retentate must be purified in a second step for a more complete separation. In gas separations, the use of retention is not practical, but selectivity is used: ci (cj) Sij =

permeate

ci (cj)

retentate

(11.15)

11.5 Concentration polarization

Fig. 11.8

313

Molecular weight cut off of two membranes.

where Sij is the selectivity of species i over species j and ci and cj are the concentrations of i and j in the permeate and retentate.

11.5 Concentration polarization The concentration gradients shown in Fig. 11.7 represent a special situation: here the flux is sufficiently small to prevent any mass transfer limitation. This is usually the case in dense, symmetrical membranes for e.g. gas separation, which are relatively thick. Concentration gradients observed in liquids flowing through porous membranes are different in two respects: • no concentration jump at the liquid-membrane interface because there is no phase transition • concentration gradients near the interface may exist because the fluxes can be relatively high. A general picture is given in Fig. 11.9. The retained solutes can accumulate at the membrane surface, where their concentration will gradually increase. Such a concentration build-up will generate a diffusive flow back to the bulk of the solution. A steady state condition is reached when the convective transport of the solute towards the membrane is equal to the sum of the permeate flow plus the diffusive back-transport of the solute: dc (11.16) = J ⋅ cP J⋅c + D dx

314

11 Membrane filtration

Fig. 11.9

Concentration polarization near a membrane.

where J is the flux through the membrane, m3/m2.hr, c the concentration in the boundary layer, kg/m3, cP the concentration in the permeate, kg/m3, D the diffusion coefficient of i, dc/dx the gradient of the concentration in the boundary layer and x the thickness of the boundary layer. The following boundary conditions apply: x = 0: c = cm x = δ: c = cb where cm is the concentration at the membrane surface and cb the concentration in the bulk. Integration of equation 11.16 results in: ln

cm − cp cb − cp

=

J⋅δ

(11.17)

D

or: cm − cp cb − cp

= exp

J⋅δ

(D )

(11.18)

The ratio of the diffusion coefficient D and the thickness of the boundary layer δ is called the mass transfer coefficient k = D/δ. The ratio of cm / cb is called the concentration polarization modulus. This ratio increases with increasing flux (cm will be higher), with increasing retention (higher cm) and with decreasing mass transfer coefficient k. When the solute is completely retained by the membrane (R = 1 and cp = 0), eq. 11.18 becomes

11.5 Concentration polarization

cm cb

= exp

J

(k )

315

(11.19)

From this formula, it is clear that the flux J and the mass transfer coefficient k are the two factors responsible for the concentration polarization. The consequences of concentration polarization can be summarized as follows: – retention can be lower Because of the increased solute concentration at the membrane surface (higher cm), the observed retention, Robs = 1 – cp / cb , will be lower than the real or intrinsic retention, Rint = 1 – cp / cm . This is generally the case with low molecular weight solutes, such as salts. – retention can be higher When macromolecular solutes are being retained, these compounds can form a secondary membrane, resulting in a higher retention of the other solutes present. – flux will be lower The flux is proportional to the driving force. When concentration polarisation and/or the formation of a secondary membrane is severe, the combined resistance of the boundary layer, the secondary membrane and the membrane itself will be high and, consequently, the flux decline can be quite considerably, whereas in gas separations concentration polarization hardly occurs and consequently, the flux remains at about the same level. Values of mass transfer coefficients, which are related to the diffusivity by the film thickness, depend on flow regime (laminar or turbulent), module geometry and fluid properties. Empirical relations between Sherwood number Sh and Reynolds number Re provide a way to estimate mass transfer coefficients. For turbulent flow in a circular tube, the mass transfer rate is characterized5 by Sh ≡

ki Dh Dim

= 0.023 × Re0.8 × Sc 0.25

(11.20)

where the hydraulic diameter Dh equals the tube diameter Dtube , Dim

= coefficient of molecular diffusion of species i, ρf v Dtube Re = ηf ηf Sc , Schmidt number = ρf Dim ηf , ρf , v = viscosity, density and velocity of the feed, respectively 5

H. Strathmann, Membranes and Membrane Separation Processes, in Ullmannʼs Encyclopedia of Industrial Chemistry (1990), Vol 16A, p. 237.

316

11 Membrane filtration

It should be noted that Eq. 11.20 applies to local flow conditions. Membranes are applied in modules (see next section) and for a proper design the flow patterns in the chosen module must be taken into account. For laminar flow in a circular tube, a similar equation is available: Sh = 1.86 × Re0.33 × Sc 0.33 ×

Dh 0.33

(L)

(11.21)

where L = length of tube From these equations, it can be seen that the mass transfer coefficient k is mainly a function of the velocity of the feed flow (v), the diffusion coefficient of the solute (D), the viscosity, the density and the module shape and dimensions. Of these parameters, the flow velocity and the diffusion coefficient are the most important. In reverse osmosis, the consequence of concentration polarization is even more severe. The osmotic pressure of the non-permeating solute is higher at the membrane surface than in the bulk, as can be seen in Figure 11.9. As a consequence, the external pressure must be (much) higher than in the absence of concentration polarization to compensate the increased osmotic pressure. This phenomena limits the applicability of RO in e.g. desalination of seawater. The osmotic pressure of seawater at several salt concentrations is tabulated in Table 11.1, together with the osmotic pressure of two other aqueous solutions. From this table, it appears that the maximum concentration in sea salt solution at the membrane interface amounts to approximately 10 wt%. Higher concentrations would require much too high a value for the external pressure to overcome the osmotic pressure.

11.6 Membrane modules So far, we discussed the performance of membranes in terms of material properties, driving forces and local mass transfer rates. Polymer membranes come usually in the form of sheets or hollow fibers, while ceramic membranes are applied as coated tubes or monoliths. A bare membrane cannot be applied without proper connections to direct a feed stream onto the membrane surface and to collect the permeate stream. The housing that enables just all that is called a module. In this section, some typical membrane modules and module flow patterns will be discussed. In Figures 11.10 to 11.12, four types of common membrane modules are shown. A plate-and-frame module, (Fig. 11.10) contains flat membrane sheets, supported on plates which direct the permeate flow. Alternatively, two membrane sheets, separated by a porous spacer, can be wound around a perforated tube to form a spiral wound module, often used in reverse osmosis for the production of potable water, see Fig. 11.10 and 11.11. The permeate flows through the

11.6 Membrane modules

317

Fig. 11.10 Flat sheet and spiral wound modules.

Fig. 11.11

Spiral wound module. (Reproduced with permission from [15].)

porous spacer to the central collection tube. In such a module the packing density is usually higher than in a plate-and-frame module. Ceramic membranes are often applied as porous tubes. A tubular module is shown in Fig. 11.12. Due to the dimensions of ceramic tubes, just a limited number can be packed in a module. Polymer hollow fibers are much thinner and the resulting packing density in a hollow fiber module is very high. The main characteristics and applicability of these four types are summarized in Table 11.2 and in Table 11.3.

318

11 Membrane filtration

Fig. 11.12

Tubular and hollow fiber modules.

Table 11.2

Some characteristics of membrane modules.

Area/volume Fouling Prefiltration Manufacturing Area of standm2/m3 tendency costs, $/m2 ard module, m2 Plate and frame Spiral wound Tube > 5 mm Capillary 0.5–5 mm Fiber < 0.5 mm

Table 11.3

100–500 300–1,000 100–500 500–4,000 4,000–30,000

little moderate low moderate high

10–25 10–25 none 10–25 5–10

μm μm μm μm

50–200 10–50 50–200 5–50 2–10

5–10 20–40 5–10 50–150 300–600

Applicability of membrane modules.

MF UF NF RO Gas separation PV/ VP ED

Tubular

Plate-frame

++ ++ ++ +

+ + + +

++

++ ++

Spiral wound

Capillary

+ ++ ++ ++ +

+ ++ ++ + + +

Hollow fiber

+ ++ ++

Unlike the size of distillation columns (or absorption or liquid-liquid extraction columns for that matter), the size of membrane modules is subject to serious limitations. Given a required capacity, the feed and product composition of a certain membrane separation and the driving force, the necessary surface area can be calculated for a chosen membrane type. Usually, however, it is not possible to put the required amount of membrane into just one module. Thus, a number of similar modules have to be build, each of which contains only a fraction of the

11.6 Membrane modules

319

total surface area. Therefore, capacity increase in membrane separations does not follow the well-known 0.6 law, common in many other chemical engineering processes. A first estimate of the capacity of a single membrane module results from application of Eq. 11.3a: the flux is calculated from permeability, membrane thickness and driving force. By definition, the total permeate flow follows from the flux and the membrane surface area contained in the module. Two complications should be considered. Firstly, in a proper approach attention should be paid to the possible influence of mass transport limitations. Mass transfer takes place at either surface of a membrane and at high overall permeation rates either side can be subject to mass transport limitation, depending on geometry. Thus, it is appropriate to look at possible combinations of flow patterns that govern the mass transfer coefficients and compare the expression for Sherwood numbers in Eqs. 11.20 and 11.21. Secondly, the geometry and, in connection with that, flow patterns in a module play an important role, because they determine the driving force as a function of position. Four important combinations of flow patterns can be distinguished: • • • •

cocurrent flow countercurrent flow crossflow perfect mixing at both sides

These patterns are depicted in Fig. 11.13. It is usually far from straightforward to identify the flow patterns in the module types mentioned above. For instance, the

Fig. 11.13 Important flow patterns in membrane modules: (a) cocurrent; (b) countercurrent; (c) crossflow; (d) perfect mixing at both sides.

320

11 Membrane filtration

flow pattern in a hollow fiber module can be approximated as counter current, but co current flow or cross flow are possible as well. At high permeate rates in a spiral wound module, the permeate moves away from the membrane surface, characteristic of cross flow. At low permeate rates, on the other hand, the flow type may be described as counter current. Here, the design of a membrane module is limited to the situation where the driving force is constant throughout the module: perfect mixing at both sides. In many cases this simplification will give a reasonable estimate. More complicated situations are outside the scope of this textbook. Interested readers are referred to specialised literature, see references and further reading. The modelling of a simple membrane separator in which perfect mixing can be assumed at both sides of the membrane is illustrated with the separation of air into an oxygen-rich permeate and a retentate that is enriched in nitrogen (oxygen is the faster permeating component). Air is considered a binary mixture of ideal gases and the module (housing, gas phase and membrane) is kept at a constant temperature.

11.6.1 Design procedure Given the feed flow F, the pressure p at both sides, the molar fraction of oxygen xF in the feed and the permeance of the chosen membrane for oxygen and nitrogen, how to calculate the unknowns such as oxygen concentration in the permeate, the recovery of oxygen (the fraction of oxygen in the permeate relative to the amount of oxygen in the feed)?

11.6.2 Solution Like any other unit operation, the model is based on material balances. In this case the overall and component balances are: F=P+R

(11.22)

xF F = yP P + xR R

(11.23)

and

where F, P and R indicate the molar flows of the feed, permeate and retentate, respectively; the indices refer to the same flows. The transport of oxygen and nitrogen through the membrane follows from the definition equations of permeance for a pressure driven process: yP P = AM

PM, O 2 δM

(xR pR − yP pP)

(11.24)

11.7 Concluding remarks

(1 − yP) P = AM

PM, N2 δM

[(1

− xR) pR − (1 − yP) pP]

321

(11.25)

where AM is the membrane surface area and PM,i / δM the permeance of component i. So far we have 11 variables (counting the permeance of a component for 1) of which 6 are given, and only 4 equations. An additional relationship or value is required to be able to solve the set of equations. A possibility is to lock, e.g., the permeate concentration. In distillation calculations the reflux offers the additional equation. In membrane calculations often the cut, the part of the feed that is allowed to permeate, is fixed: cut =

P F

(11.26)

Now the 5 equations Eqs. 11.22–11.26 can be solved. An efficient route to do this is to firstly calculate P and R from Eqs. 11.22 and 11.26. Secondly, eliminate the membrane surface area from Eqs. 11.24 and 11.25 and replace the ratio of permeances by the selectivity (see section 11.4). Then the molar fractions yP and xR can be solved from Eqs. 11.23 and 11.24. Subsequently, the eliminated membrane surface area can be calculated from Eq. 11.24. Finally, from the values thus obtained, any parameter of interest, such as the separation factor or the oxygen recovery, can be calculated. Though an analytical solution exists, an example calculation is given on page 323 of this chapter. It shows a numerical solution as programmed in MathCad. As a final result it is shown how the required membrane surface area varies with a chosen value of the cut. Obviously, the higher the cut, the higher the required surface area.

11.7 Concluding remarks Membrane processes possess a number of distinct advantages compared to alternative separation methods, such as distillation, absorption and extraction. The usefulness of membrane applications may emerge from the following: • Numerous separations can be carried out with the use of membranes because of the very broad scale of particles sizes covered, as shown in Fig. 11.4. • Energy requirements of membrane separations are usually low because generally no phase change is involved. An exception is pervaporation, where the feed is in the liquid phase and the permeate in the vapour phase. • Membrane separations usually function effectively at room temperature. • The list of suitable membrane materials is long and still expanding. This means that in many cases appropriate membranes exist with a high selectivity for the species to be separated.

322

11 Membrane filtration

• A membrane unit simply comprises a number of membrane modules in parallel and a suitable number of compressors (or pumps). The above may give the impression that a membrane process is the feasible technology for almost any separation. However, the use of membranes for separation is seriously limited due to the following disadvantages: • Usually, membrane processes have only a few stages, which requires high selectivities, much higher than would be necessary in e.g. distillation where a lower selectivity (relative volatility) is compensated by a higher number of stages. • Chemical resistance can be an issue. Polymer membranes may weaken, swell, or even dissolve in contact with a variety of organic compounds, such as aromatics and oxygenated solvents. • Sometimes it is impossible to apply membranes because of lack of thermal stability above the required temperature (dictated by compatibility with connected unit operations). Not only the (polymer) membrane cannot withstand temperatures much above say 80 °C, the same is true for glues and other sealing material normally used in membrane modules. Ceramic-based membranes, on the contrary, can be used to much higher temperatures. However, sealing problems due to differences in thermal expansion in the latter modules prevent their large-scale application, at the time being. • Scaling up membrane units means simply applying more modules in parallel in stead of installing bigger modules. Therefore, the investment of a membrane separation process will increase almost linearly with scale and may be favourable only below a certain maximum capacity as compared to unit operations obeying the power 0.6 law, such as distillation, extraction and absorption. • Fouling of membranes may greatly restrict permeability of membranes and make them even unsuited for some types of feed streams or expensive regeneration schemes are necessary. Also longer-term flux decline may occur, limiting the membraneʼs lifetime to only a fraction of the expected time-on-stream. • In some applications, especially at high osmotic pressure, the cost of compression energy may be high. In conclusion: membrane separations have found their way into the chemical industry and its market is still expanding because of ongoing improvement of membrane materials and membrane module design.

11.7 Concluding remarks

323

Calculation of the required membrane area as a function of the cut and feed pressure F := 50

xF := 0.2

PO2 := 0.04 PN2 :=

feedlow [mol/s] and mole fraction O2 in feed

PO2

membrane properties, permeances [mol/(s.m2.bar)]

4

+

pP := 1.5

pressure product side [bar]

yP := xF

xR := xF

initial guesses for solving Eq. 11.19–11.23

given

(1 – yP) ⋅ PO2 ⋅ (xR ⋅ pR – yP ⋅ pP) = yP ⋅ PN2 ⋅ [pR ⋅ (1 – xR) – pP ⋅ (1 – yP)] xF ⋅ F = yP ⋅ cut ⋅ F + xR ⋅ (1 – cut) ⋅ F solve_for (cut, F, pR) := Find (yP, xR)

j := 4 .. 6

pRj := j

i := 0 .. 20

cuti := 0.1 +

yPi

[xRi]

three feed (= retentate) pressures [bar] i 25

:= solve_for (cuti, F, pRj)

AM (i, pRj) yPi ⋅ cuti ⋅ F := PO2 ⋅ (xRi ⋅ pRj − yPi ⋅ pP)

range of cuts calculated product and retentate mole fractions membrane area AM as a function of the cut for feed pressures of 4, 5 and 6 bar

3000 AM (i,4) AM (i,5) 2000

AM (i,6)

1000

0 0

0.2

0.4

0.6 cut i

Fig. 11.14

0.8

1

324

11 Membrane filtration

Nomenclature D F, P, R Ji Ki p PM,i Ri v¯ x, y Sij δM η ρ λ

coefficient of diffusion feed, product and retentate flow component flux distribution coefficient of component i pressure permeability of component i, see Eqs. 11.3a and 11.3c retention of i, Eq. 11.14 mean speed mole fractions selectivity, Eq. 11.15 membrane thickness viscosity density mean free path

[m2 s–1] [mol s–1] [mol s–1 m–2] [–] [N m–2] [–] [m s–1] [–] [–] [m] [Pa.s] [kg m–3] [m]

Exercises 1 Show that permeability can be expressed in m2 s–1. 2 The concentration of a solute at both sides of a membrane with thickness 30 μm is 0.030 kmol/m–3 and 0.0050 kmol/m–3. The equilibrium constant K = 1.5 and DA = 7.0 10–11 m2 s–1 in the membrane. The mass transfer coefficients at both liquid sides can be considered as infinite. Calculate the flux and the concentrations at the membrane interfaces inside the membrane. 3 Repeat the previous exercise in case that the mass transfer coefficient at the product side amounts to 2 10–5 m s–1. 4 Calculate the osmotic pressure of a solution containing 0.1 mol NaCl/kg H2O at 25 °C. The density of pure water at 25 °C amounts to 997.0 kg/m–3. Compare the result with the experimental value given in Table 11.1. 5 The permeability of a membrane is experimentally determined by measuring the flow of an aqueous NaCl solution in a reverse osmosis module at a given pressure difference. The inlet concentration cin = 10 kg /m3 solution with a density of 1004 kg/m3 solution. The product concentration cout = 0.39 kg/m3 solution with a density of 997 kg/m3 solution. The measured flow rate is 1.92 10–8 m3 solution/s at a pressure difference of 54.4 atm, the membrane area A = 0.0020 m2. Calculate the permeability constants for water and sodium chloride.

12 Separation method selection

12.1 Introduction The aim of this chapter is to explore the issue of how to choose the right separation process, or narrow down the number of choices for a given separation. In most of the cases, however, this selection process is complicated because there is no standard procedure by which separation processes can be chosen. Therefore we will follow a heuristic approach using simple rules of thumb and direction guidelines that can be followed to arrive at the separation process sequence to be adopted in the final plant design. The degree to which this approach reaches the optimum separation scheme depends on the time and money allocated to the development and analysis and on the skills of the process design engineer in applying guidelines to the selection of sequences for detailed evaluation. When the feed to a separation system is a binary mixture, it may be possible to select a separation method that can accomplish the separation task in just one piece of equipment. In that case the separation system is relatively simple. In most practical situations, however, the feed mixture involves a complex mixture of many components. For these mixtures a separation train is required consisting of a number of operations in which the separations are sequenced. The separation in each operation (unit) is made between two components designated as key components for that particular separation unit. Each effluent is either a final product or a feed to another separation device. The synthesis of a multicomponent separation system can be very complex because it involves not only the selection of the separation methods, but also the manner in which the separation operations are sequenced. We will then deal with how a series of separations should be oriented to separate a complex mixture. 12.1.1 Industrial separation processes The complexity of recovering products in industrial chemical processes is clearly illustrated by Fig. 12.1 in which the flow sheet of a naphtha cracker is shown. In the reactor furnaces naphtha is cracked at a temperature of around 900 °C to a mixture of hundreds of light hydrocarbons, of which the unsaturated hydrocarbons (ethylene, propylene, butylene) are the main products. To recover these products the gaseous reactor effluent is first quenched in an absorption tower with heavy oil to cool the gas stream and condense the heavy oil constituents. In a subsequent operation some acid gases are removed by reactive absorption to prevent corrosion

326

12 Separation method selection

Fig. 12.1 Schematic flow sheet of a naphtha steam cracker.

problems in the final distillation unit. Further separation is accomplished in a series of distillations in which first the C2, C3 and C4 hydrocarbons are recovered and finally the saturated and unsaturated hydrocarbons are separated. Traces of acetylene are removed from the C2 hydrocarbon stream by selective absorption in DMF prior to the final distillation. A second example of a process with a complicated separation train to recover the final product from the reaction mixture is the industrial synthesis of caprolactam. The flowsheet of the final step in this process is schematically represented in Fig. 12.2. After the reaction in a strongly acidic medium the caprolactam is obtained by neutralization with ammonia, resulting in a two-phase mixture consisting of an ammonium sulphate and a caprolactam rich phase. The caprolactam is recovered from the top layer by extraction with an organic solvent such as benzene, toluene or cyclohexane. A second extraction is used to recover residual caprolactam from the ammonium sulphate rich phase. The caprolactam is recovered from the solvent by back extraction into water and further purified by various techniques such as ion exchange to remove residual contaminants before final evaporation of the water. For both processes it is interesting to analyze why the used separation methods have been selected and how it would be possible to improve the process concept by the selection of different separation methods. It is clear that the compositions

12.1 Introduction

Fig. 12.2

327

Schematic representation of the final step of the caprolactam synthesis.

of the mixtures to separate, scale of operation and the minimization of unnecessary phase changes are important factors that have to be considered. For instance, in a naphtha cracker absorption is used to selectively remove trace amounts of acid gases without the need of a phase change. Subsequent recovery and purification is done by distillation because this is the most economical large-scale separation technology. The most difficult separation is performed last, being the ethylene/ethane and propylene/propane separations. In the case of caprolactam, extraction is employed because the product has a boiling point in between the water and residual ammonium sulphate. Back-extraction is used to remove non-polar impurities that were coextracted with caprolactam in the forward-extraction step. 12.1.2 Factors influencing the choice of a separation process 12.1.2.1 Economics In most situations the ultimate criterion for the selection of a separation method is economics. Determining factors are the economic value of the products and the

328

12 Separation method selection

scale of operation. However, the economic criterion is usually subject to a number of constraints originating from a corporate attitude such as market strategy and timing, reliability, risks associated with innovation, and capital allocation. Two extreme cases are very helpful to illustrate the influence of these constraints: 1. The product is of high unit value with a short market life expectancy 2. The product is a high-volume chemical with many producers in a highly competitive market In Case 1 the procedure would probably be to choose the first successful separation method found. The ultimate overall economics of this situation are determined by getting into the market ahead of the competition and selling as much of the product as possible while the market lasts. Clearly many separation processes, which would be unsuitable for substances with a low value, can be considered for the recovery and purification of these high value products. In Case 2, the ultimate viability of the plant dictates that the process development and design teams do the most thorough evaluation of various process schemes possible to make the closest approach to the economic optimum design. The desired plant capacity can be an important factor in separation process selection, since some processes are difficult to carry out on a very large scale. The lower the economic value of the product, the more important it will be to select a process with relatively low energy consumption and to select a process where the unit cost of any added mass separating agent is relatively low. A small loss of a costly mass separating agent can be an important economic penalty to a process. 12.1.2.2 Feasibility First and foremost, any separation process to be considered in a given situation must have the potential of giving the desired result. This feasibility criterion is commonly used to make a first selection between separation methods that may work and certainly wonʼt work. Often the question of process feasibility will have to do with the need for extreme processing conditions. Although the dividing lines are not easy to draw, the general idea is that a process which requires very high or very low pressures or temperatures will always suffer in comparison with one that does not require extreme conditions. Another situation where the feasibility condition enters into consideration occurs when a mixture of many components is to be separated into relatively few different products. In such a case it will be necessary for each component to enter the proper product. Since different separation processes accomplish the separation on the basis of different principles, it is quite possible for different processes to divide the various components in different orders between products.

12.1 Introduction

329

12.1.2.3 Product stability Often the question of avoiding damage to the product can be a major consideration in the selection of a separation process. Thermal damage may be manifested through denaturation, formation of unwanted color, polymerization etc. When thermal damage is a factor in separation by distillation, a common approach is to carry out the distillation under vacuum to keep the reboiler temperature as low as possible. Frequently evaporators and reboilers are given a special design to minimize the holdup time at high temperature of material passing through them. 12.1.2.4 Design reliability Of all factors influencing the decision to choose one process in preference to another, design reliability is the most important factor. Regardless of any other considerations, the constructed plant must work properly to produce an acceptable product that can be sold at a profit. The economic consequences of a design failure are too dramatic to accept any potentially malfunctioning separation process design. It is therefore not surprising that those separation operations that are well understood and can be readily designed from first principles are favored in an industrial environment. This is illustrated by Fig. 12.3, showing that the degree to which a separation operation is technologically mature correlates well with its commercial use. Design reliability is not really definable in quantitative terms because it actually relates to the amount of testing and demonstration that must be done before a suitable commercial scale is produced.

Fig. 12.3

Technological and use maturities of separation processes.

330

12 Separation method selection

12.2 Selection of feasible separation processes 12.2.1 Classes of processes In the process of selecting a feasible separation process some generalizations can be drawn with respect to the advantages and disadvantages of different classes of separation processes for various processing applications. Important in this classification is the distinction between energy-separating-agent, mass-separating-agent and rate-governed processes. In general the energy consumption of the process increases as we go from energy separating agent to mass separating agent to rate-governed processes. Therefore a multistage rate-governed separation process should only be considered when it gives a significantly better separation factor than an equilibrium based process. In the same way a mass-separating-agent process should only be considered when it provides a better separation factor than an energy-separating-agent process. The higher energy consumption of the massseparating agent process for a given separation factor is associated with the introduction of an additional component and the need for removing this component from at least one of the products. 12.2.2 Initial screening The selection of a best separation process must frequently be made from a number of candidates. When the feed mixture is to be separated into more than two products, a combination of two or more operations may be the best. Even when only two products are to be produced, a hybrid process of more than two or more operations may be required or the most economical. The first step in the selection of a separation process is to define the problem. Table 12.1 gives an overview of four important categories that have to be considered in the definition of the separation problem and the selection of feasible separation operations. These factors have to do with feed and product conditions, property differences that can be exploited and certain characteristics of the candidate separation operations. The most important feed conditions are composition and flow rate, because the other conditions can be altered to fit the required conditions of a candidate separation operation. In general, however, the vaporization of a liquid feed that has a high heat of vaporization, the condensation of a vapor feed with a refrigerant, and/or the compression of a vapor feed can add significantly to the cost. The most important product conditions are the required purities and recovery specifications because the other conditions can be altered by energy transfer after the separation is achieved. Product purity specifications should ultimately be established by customers, although they may not be stated explicitly. Recovery specifications are set to assure an economic process. Ideally, the recovery specification should be a variable to be optimized by the process design. In practice, there is usually insufficient time to consider recovery as a variable, and it is specified either arbitrarily or by consensus.

12.2 Selection of feasible separation processes

331

Table 12.1 Factors that influence the selection of feasible separation operations.

Feed Conditions

Product Conditions

Property differences to be exploited

Characteristics of separation operations

Composition, particularly concentration of species to be recovered or separated Flow rate Temperature and Pressure Phase state (solid, liquid, gas, mixed) Required product purities Required product recovery Temperature and pressure Phase states Molecular Thermodynamic Transport Ease of scale-up Ease of staging Temperature, pressure and phase-state requirements Physical size limitations Energy requirements

As mentioned before some separation operations are well understood and can be readily designed and scaled up to a commercial size from laboratory data. Table 12.2 ranks the more common separation operations according to ease of scale-up. Operations ranked near the top are frequently designed without the need for any laboratory data or pilot-plant tests. Operations near the middle usually require laboratory data, while operations near the bottom require pilot plant testing on actual feed mixtures. Also included in the table is an indication of the ease of providing multiple stages and to what extent parallel units may be required to handle high capacities. Single-stage operations are utilized only when the separation factor is very large or only a rough or partial separation is needed. When higher product purities are required, either a large difference in certain properties

Table 12.2 Ease of scale-up and staging of the most common separation operations.

Operation in Decreasing Ease of Scale Up

Ease of Staging

Need for Parallel Units

Distillation Absorption Extractive and Azeotropic Distillation Liquid-liquid Extraction Membranes Adsorption Crystallization Drying

Easy Easy Easy

No need No need No need

Easy Repressurization between stages Easy Not Easy Not Convenient

Sometimes Almost always Only for regeneration Sometimes Sometimes

332

12 Separation method selection

must exist or efficient countercurrent-flow cascades of many contacting stages must be provided. Operations based on a barrier are generally more expensive to stage than those based on the use of a solid agent or the creation or addition of a second phase. Some operations are limited to a maximum size. For capacities requiring a larger size, parallel units must be provided. 12.2.3 Separation factor An important consideration in the selection of feasible separation methods is the separation factor that can be achieved for the separation between two key components of the feed. This factor has already been extensively discussed in chapter 1 and is defined for the separation of component 1 from component 2 between phases I and II in a single stage of contacting as (see also Eq. 1.1): C1I C1II SF1,2 =

C2I

(12.1)

C2II where C is a composition variable, such as mole fraction, mass fraction or concentration. If phase I is to be rich in component 1 and phase II is to be rich in component 2, the separation factor is large. The value of the separation factor is limited by thermodynamic equilibrium, except in the case of membrane separations that are controlled by relative rates of mass transfer through the membrane. In general components 1 and 2 are designated in such a manner that the separation factor is larger than unity. Consequently, the larger the value of the separation factor, the more feasible is the particular separation operation. In general, operations employing an energy-separating agent are economically feasible at a lower value of SF than those employing a mass-separating agent. This is illustrated in Fig. 12.4, where the required separation factor for considering extractive distillation and extraction are plotted against the separation factor obtained in classical distillation. In general mass separating agents for extractive distillation and liquid-liquid extraction are selected according to their ease of recovery for recycle and to achieve relatively large values of the separation factor. The objective of the design of a separation process is to exploit the property differences in the most economical manner to obtain the required separation factors. Processes that emphasize molecular properties in which the components differ to the greatest should be given special attention. This will be illustrated by several examples in the following paragraphs.

12.3 Separation of homogeneous liquid mixtures

333

Fig. 12.4 Relative separation factors for equal-cost separation operations. (Adapted from M. Souders, Chem. Eng. Prog., 60(1964)2 75–82)

12.3 Separation of homogeneous liquid mixtures At present distillation is used for over 90 % of all separations in the chemical process industry. Taking cognizance of this dominance, distillation, although not energy-efficient, is apparently the benchmark against which all newer processes must be compared. Therefore we will start with an analysis why distillation is so dominant and then proceed to define those separation conditions where the selection of other processes is favored. Finally the conditions will be defined where one process is preferred over another. 12.3.1 Strengths of distillation The comparisons between the different classes of separation processes so far lead rather strongly toward distillation. Economically, if a stream can be easily vaporized or condensed, distillation or a related vapor-liquid separation process is most often the process of choice for separating the components in that stream. There are five important reasons for this dominance, which are summarized below. • Small equipment requirements Compared to virtually all other processes the equipment requirements of distillation are small. As illustrated by Fig. 12.5, usually just a column, a reboiler, a condenser and some relatively small ancillary equipment are required. Alternative processes based on mass separating agents require additional investment for the circulation and recovery of the mass-separating agent. The same holds when distillation is compared to crystallization, another energy-separating-

334

12 Separation method selection

Fig. 12.5

Schematic of a distillation column equipped with sieve trays.

agent equilibration process, involving solid phases and therefore requiring additional operations (filtration, washing, drying) to obtain the final product. All of these complications often lead to relatively high investments for alternative processes unless separation by distillation is very difficult to effect • Easy staging Distillation can be staged very easily. It is possible to have 100 or more theoretical stages in a distillation column. This means that, even at low relative volatilities down to 1.2, full separations can be achieved. In addition, mass transfer rates between phases are generally high enough to prevent excessive equipment sizes for high degrees of separation. • Economics of scale The economics of scale favor distillation for large-scale separations. Distillation technology scales up by roughly a 0.6 factor on column throughput, which means that doubling a columnʼs capacity increases the column investment only by a factor of about 1.5. I I0

=

C

0.6

(C0)

(12.2)

12.3 Separation of homogeneous liquid mixtures

335

For most alternative processes the economic improvement of going to very large scales is less pronounced because of their higher scale-up factors. This lowered potential for cost reduction is caused by the more complex flowsheets of the alternative processes and in a number of cases the necessity to place equipment in parallel to accept the whole feed stream. Membrane modules are an excellent example of this problem. To accept large flows, many modules must be aligned in parallel, and a considerable part of the savings of going to a larger scale is lost. • Energy costs In most operations energy costs are not a dominant consideration. Although this might seem a surprising conclusion at first, energy costs for distillation virtually always run second in impact to capital costs. Therefore the substitution of an existing distillation process by a more complex, higher-investment but lower-energy-usage process is often difficult to motivate on economic grounds. In addition, in some plants low-pressure steam with little or no value is generated, making energy costs very low. Heat integration of distillation columns can also be practiced on very large scales to help reduce energy costs. Two major examples are olefin plant cryogenic separation systems and cryogenic air separation. • Design and scale-up reliability Distillation is the only separation process that can be designed from the physical properties of the components and vapor-liquid information on the important binary-pair components alone. Reliability design methods for distillation have been developed over many years of industrial experience and extensive testing of commercial-scale equipment. Occasionally, some small-scale testing is required, but scale-up methods for distillation are the most reliable of all the separation methods. Because of this favorable combination of factors, it is no coincidence that distillation is the most frequently used separation process in practice. In fact, a sound approach to the selection of appropriate separation processes for the separation of homogeneous liquid mixtures is to begin by asking: Why not distillation? Unless there is some clear reason why distillation is not well suited, distillation will be a leading candidate. 12.3.2 Limitations of distillation In spite of this impressive array of factors, which cause distillation to be favored, other considerations have lead to the development and installation of other separation techniques in many applications. Some situations which lead to distillation separation costs which are higher than those of alternative processes, or in which other factors decide against the use of distillation:

336

12 Separation method selection

• Low relative volatility Distillation becomes very expensive when the relative volatility between key components is too low. Seldom will a relative volatility over part or the entire distillation curve of less than about 1.2 give acceptable economics for distillation. Relative volatilities less than 1.2 obviously include azeotrope-forming situations, for which the relative volatilities become 1 at the azeotropic point. In such cases azeotropic or extractive distillation, pervaporation, solvent extraction and adsorption should be investigated for providing reasonable solutions. • Feed composition Feed compositions and product purity requirements can present unusual problems. Three problems dominate this category: – If the product of interest is a high boiler in low concentration, less than 20 wt% in the feed, energy costs per unit of desired product to boil up all of the low-boiling material, as well as column capital cost, can become excessive. A typical example of this situation is the liquid phase oxidation of cyclohexane where conversions of only a few percent are used to achieve the desired selectivities. – If a small concentration (only a few percent) of high-boiling contaminants must be removed from a desired product, the energy and capital cost for distillation can also become excessive. – If the boiling range of one set of components overlaps the boiling range of another set of components from which it must be separated, an extreme amount of distillation columns would be required for a complete separation. In these situations mass-separation-agent based processes have the advantage of being able to separate one set from another in one step based on their difference in chemical structure. Industrial well known examples are olefin/paraffin and paraffin/aromatic separations. • Extreme conditions If distillation temperatures are less than about –40 °C or greater than 250 °C, the cost of refrigeration to condense column overheads in the first case and heat to vaporize in the second escalate rapidly from typical cooling water and steam costs. If required column pressures are less than about 20 mbar, then column-size and vacuum pump costs can escalate rapidly. If column pressures higher than about 50 bars are required, column investment can also escalate as a result of the thick vessel walls to meet the pressure requirements. • Small capacities Earlier it was mentioned that distillation has a relatively low scale-up factor of about 0.6. Unfortunately if a process scales up well economically, that same process does not scale down well. That is, its investment does not reduce, as throughput is reduced, as much as other separation techniques with a higher scale-up factor. Thus alternative processes, which do not compete for large production rates, may compete well at smaller production rates.

12.3 Separation of homogeneous liquid mixtures

337

• Product degradation Various types of undesirable reactions can sometimes occur at column temperatures. Thermally labile components in the feed can undergo reactions, which result in either significant product loss or in the formation of unwanted byproducts that are difficult to separate. In addition the intrusion of oxygen from small leaks under high vacuum can cause severe product purity problems through unwanted side reactions. • Column fouling Column fouling rates can be unacceptable. Although product loss rates might be completely acceptable and there are no product quality concerns, precipitation or polymerization reactions can cause unacceptable fast column fouling. Especially when a column cannot be operated for a few days before it must be shut down and cleaned will hardly be economical unless alternative processes are incredibly expensive. Despite its industrial dominance, this paragraph illustrates that distillation is not always the right answer for a given separation. The other processes we have discussed obviously have their places in the scheme of separation trains encountered in industrial processes. In fact, new separation opportunities, since they will often involve more complex and less volatile mixtures than the present spectrum of separations, will be served by these alternative processes. The following paragraphs have the objective to illustrate how alternative separation methods can be used when the conditions are unfavorable for applying distillation. Three often-encountered situations (low relative volatility, overlapping boiling points and low concentrations) will be analyzed on the basis of industrial examples to clarify the advantages of alternative technologies. Because of its exemplary nature, the approach is not to present an all-covering extensive analysis but merely to show how the exploitation of different separation principles can provide alternative solutions to separation problems. 12.3.3 Low relative volatilities Low relative volatilities are generally encountered when the difference in boiling points of components to be separated is insufficient to achieve a good separation. In principle two distinct situations exist, that both require a different approach to achieve full separation. An example of the first situation is the separation of benzene from cyclohexane or 1-butene from isobutane. A comparison of the properties of these components in Table 12.3, illustrates clearly that the difference in boiling point between cyclohexane and benzene results only in a relative volatility of 1.01, which is far too small to apply distillation. Although the boiling point difference for the 1-butene/isobutane system is larger, non-ideal behavior of the liquid phase is responsible for obtaining only very low relative volatilities. In

338

12 Separation method selection

Table 12.3 Some pure component properties of benzene, cyclohexane, 1-butene and isobutane.

Property

Benzene

Cyclohexane

1-Butene

Isobutane

Molecular weight Boiling point (°C) Melting point (°C) Dipole moment (debye) Polarizability (10–31 m3)

78.1 80.1 5.6 0.0 103

84.2 80.7 6.5 0.3 110

56.1 –6.2 –185.3 0.3 80

58.1 –11.7 –159.5 0.1 82

Fig. 12.6

Schematic of an extractive distillation process.

these cases the molecular difference in chemical structure can be exploited to increase the separation factor to a sufficiently high level. This can be done by the addition of a polar solvent (n-methylpyrrolidone, sulfolane, furfural) that increases the volatility of the hydrocarbon stronger than the volatility of the aromatic. So introducing a polar solvent in the top of a distillation tower will preferentially force the cyclohexane to the vapor phase and thereby facilitate separation. In most of these low relative volatility situations azeotropic and extractive distillation are economically favored because of their high mass transfer rates and ease of phase disengagement. As illustrated by Fig. 12.6, a second distillation column is re-

12.3 Separation of homogeneous liquid mixtures

339

Table 12.4 Some pure component properties of xylenes and ethylbenzene.

Property

o-Xylene

m-Xylene

p-Xylene

Ethylbenzene

Molecular weight Boiling point (°C) Melting point (°C) Dipole moment (debye) Polarizability (10–31 m3) Molecular volume (cm3/mol) Kinetic diameter (Å)

106.2 144.2 – 25.0 0.21 141 121

106.2 139.5 – 47.7 0.10 142 123

106.2 138.7 13.5 0.00 142 124

106.2 136.5 – 94.7 0.20 135 123

6.8

6.8

5.8

6.3

quired to separate the benzene from the extractive solvent. The resulting flow sheet is indeed more complex than a distillation flow sheet, but the mass transfer rate and the phase disengagement ease of distillation are also present, making them economically favored in a number of low relative volatility situations. An example of the second situation is the separation of xylene isomers and ethylbenzene. Xylenes are obtained commercially from the mixed hydrocarbon stream manufactured in naphtha reforming units in oil refineries. p-Xylene is the desired product for the manufacture of phthalic acid and dimethyl terephthalate. As shown in Table 12.4, the boiling points of p-xylene, m-xylene and ethylben-

Fig. 12.7 tion.

Schematic process for p-xylene manufacture by crystallization and isomerisa-

340

12 Separation method selection

zene differ only marginally. With a separation factor of only 1.02 ordinary distillation would require about 1000 stages and a reflux ratio of more than 100. Because also the chemical nature of all four constituents is almost the same, the addition of an extractive solvent will act approximately the same on all isomers. The only macroscopic property that differs considerably is the melting point, allowing selective recovery of the p-xylene from the mixture by crystallization. For crystallization the separation factor is nearly infinity due to the additional advantage of shape selectivity in crystal growth. The resulting flowsheet for industrial p-xylene recovery is shown in Fig. 12.7. Other properties that may be successfully exploited in the separation of isomer mixtures are the difference in kinetic diameter and complexation behavior resulting from the difference in isomer structures. In the case of p-xylene separation this is used in a large-scale continuous adsorption process where shape selective zeolites are used as adsorbents to achieve full separation. One of the properties they use is the relative linearity of the p-xylene molecule compared to the other xylene isomers. As a result the p-xylene molecules fit better, the kinetic diameter is smaller, into the zeolite pores and an increase in the separation factor to values exceeding 3 is obtained. 12.3.4 Overlapping boiling points Overlapping boiling points are generally encountered when it is attempted to separate one class of components from another class. Industrial examples are the alifatics/aromatics, olefins/paraffins and paraffins/isoparaffins separation. In these cases the strategy for obtaining a suitable separation factor is roughly the same as with low relative volatilities. The main difference however, is that vapor-liquid equilibria are avoided because it is almost impossible to overcome the large differences in vapor pressure by enhancement of the separation factor. In the case of alifatics/aromatics the desired separation factor is obtained by introducing a highly-polar solvent to form a non-ideal solution that will split into two phases. The used solvents such as sulfolane, n-methylpyrrolidone and tetraethylene glycol have a purely physical interaction with the hydrocarbon phase. The alifatics are repelled stronger by the polar solvent than the aromatics, generating the desired separation factor. Unique in these processes is that the extraction columns are operated with a partial recycle of the extract in order to obtain full separation of the alifatics and aromatics. As shown in Fig. 12.8, both product streams extract and raffinate have to be distilled to recover the solvent. These additional distillations make an extraction process in general economically unattractive, except for situations where a large amount of distillation operations would be required to obtain the same separation.

12.3 Separation of homogeneous liquid mixtures

Fig. 12.8

341

Schematic of extraction processes for alifatics/aromatics separation.

12.3.5 Low concentrations If only a small amount of one component is to be removed from a large amount of one or more other components, changing the phase of the latter components should be avoided. In such a case, stripping, extraction or selective adsorption are the preferred techniques to remove the minor component. Stripping is often used to remove small amounts of volatile chemicals from aqueous systems. It is particularly good for components that are immiscible with water and tend to have extremely high volatilities such as benzene, toluene and other hydrocarbons. Stripping of low volatility components that are completely miscible with water is generally uneconomical because of the large amounts of stripping gas or steam that are required. Extraction is generally favored for the removal or recovery of reasonable amounts of a desired component with a low volatility from reaction mixtures or waste streams. Such examples are often encountered in biotechnological processes where the product stream contains typically up to 10 wt% of the desired product (organic acids, amino acids, penicillins, alcohols), which is recovered by extraction with solvents such as heptane, methylethylketone. At very low concentrations (< 1000 ppm) extraction usually becomes highly uneconomical because the removal of solvent residues from the raffinate and

342

12 Separation method selection

recovery of solvent from the extract stream becomes too expensive. The same holds for stripping because the stripped components have to be removed from the gaseous stream. In those cases adsorption is often the technology of choice. Typical examples are the removal of contaminants by adsorption on activated coal or the removal of trace amounts of water to prevent catalyst deactivation from monomer and solvent streams in polymer processes. 12.3.6 Dissolved solids The final product from many industrial chemical processes is a solid material. For these systems of dissolved solids, such as inorganic salts in water or essentially non-volatile or high melting organics, crystallization is the predominant separation method. Even when the final product is not a solid, solid-liquid or solid-gas separation operations may be involved. An already discussed example is the separation of xylene isomers. Crystallization is induced commonly by one of four mechanisms: cooling, solvent evaporation, reaction and the addition of an anti-solvent. Cooling is the preferred mode for systems in which solubility shows a large decrease with decreasing or increasing temperature. Evaporation of solvent is the preferred mode when solubility shows little or no temperature dependence. A common flowsheet for the separation section of a process for the manufacture of solid products is shown in Fig. 12.9. In these processes a crystallizer is used to produce the solid product, always followed by a solid-liquid phase separation to remove as much of the solvent as possible without the need of evaporation. In many crystallization processes a hydrocyclone is used for control of the particle size by separation of fines that are recycled to the crystallizer. The remaining solvent residues are finally removed in a drying step. The physical nature of the material to be handled is the primary item for consideration in dryer selection. A

Fig. 12.9 Schematic of a general solids production process.

12.4 Separaton systems for gas mixtures

343

slurry will demand a different type of dryer from that required by a coarse crystalline solid. As discussed before, crystallization is an alternative to other separation techniques for liquid mixtures. However, it should always be kept in mind that separation processes involving handling of a solid have a disadvantage in continuous operation relative to processes in which all phases are fluid. This disadvantage stems from the difficulty of handling solids in continuous flow. Therefore crystallization should only be considered when other techniques are not feasible or when the desired final state of the product is solid and a crystallization operation is required anyhow.

12.4 Separaton systems for gas mixtures 12.4.1 General selection considerations In the previous section we have primarily dealt with the separation of liquid mixture feeds. The primary techniques are ordinary and enhanced distillation. If the feed consists of a vapor mixture in equilibrium with a liquid mixture, the same techniques can often be employed. If the feed is a gas mixture and a wide gap in volatility exists between two groups of chemicals in the mixture, it is often preferable to partially condense the mixture, separate the phases, and send the liquid and gas phases to separate separation systems. Whereas ordinary distillation is the dominant method for the separation of liquid mixtures, no method is dominant for gas mixtures. The separation of gas mixtures is further complicated by the fact that whereas most liquid mixtures are separated into nearly pure components, the separation of gas mixtures falls into the following three categories: 1. Enrichment to increase the concentration of one or more species 2. Sharp splits to produce nearly pure products 3. Purification to remove one or more low-concentration impurities Enrichment is defined as a separation process that results in the increase in concentration of one or more species in one product stream and the depletion of the same species in the other product stream. Neither high purity nor high recovery of any component is achieved. Gas enrichment can be accomplished with a wide variety of separation methods such as physical absorption, adsorption, condensation and membrane permeation. Sharp separations are often referred to as bulk separation and are designed to produce two high-purity product streams at high recovery. Separations in this category can be difficult to achieve for gas mixtures. The separation methods that can potentially obtain a sharp separation in a single step are physical absorption, adsorption and cryogenic distillation. Chemical adsorption is often used to

344

Fig. 12.10 ration.

12 Separation method selection

Comparison of the economics of operation of various techniques for air sepa-

achieve sharp separations, but generally limited to situations where components to be removed are present in low concentrations. The special case involving the removal of a low (< 5 mol%) mole fraction impurity at high (> 99 %) recovery is called purification separation. Purification typically results in one product of very high purity. It may or may not be desirable to recover the impurity in the other product. The separation methods applicable to purification separation include adsorption, chemical absorption and catalytic conversion. Physical absorption is not included in this list as this method typically cannot achieve high purities. Besides the separation category and separation factor, the production scale of the process is a major factor in determining the optimal separation method for the separation of gas mixtures because economies of scale are most pronounced for cryogenic distillation and absorption. This effect of economy of scale is perfectly illustrated in air separation. As Fig. 12.10 shows, membrane separations are most economical at low production rates, adsorption at moderate rates and cryogenic distillation at large production scales. 12.4.2 Comparison of gas separation techniques 12.4.2.1 Absorption Physical absorption is typically used for the recovery of components that are present in reasonable concentrations. The absorption liquid should be selected to have a separation factor of 3 or higher for economical operation. Unless values

12.4 Separaton systems for gas mixtures

345

of the separation factor are about 10 or higher, absorption and stripping operations cannot achieve sharp separation between two components. Nevertheless, these operations are used widely for preliminary or partial separations where the separation of one key component may be sharp, but only partial separation of the other key component is required. If the composition of the absorbate in the gas is to be reduced below 100 ppm in the lean gas product it should be considered to combine physical absorption with chemical absorption or adsorption. An important advantage of absorption is its favorable scale-factor, which is comparable to distillation. Therefore absorption is particularly suited for large-scale applications and situations where the product should be obtained in the liquid phase for further processing. Typical examples are the recovery of oxidation products from partially converted gas streams and the absorption of NOx and SO2 for the production of nitric acid and sulfuric acid. In these acid production applications absorption in a liquid is essential to obtain the desired product state and has the additional advantage of allowing interstage cooling in the absorption tower. When considering chemical absorption the impurities must contain acid-base functionalities or other reactive groups. The reactivities of the impurities must be much higher than the bulk stream components to allow selective impurity removal. Chemical absorption is typically favored when the desired removal is high and the partial pressure of the components to be removed from the bulk stream is low (< 1 %). One of the most widely used applications of chemical absorption is the removal of acid gases (CO2 and H2S) with amine group containing reactive solvents. 12.4.2.2 Adsorption Adsorption should be favored for processes that require essentially complete removal of water vapor and for small-scale desiccation operations. Adsorption is a very versatile separation technique that can be used for all three separation categories. Selectivity in adsorption is controlled by molecular sieving or adsorption equilibrium. When solutes differ significantly in molecular size and/or shape, zeolites and carbon molecular-sieve adsorbents can be used to advantage. These adsorbents have very narrow pore-size distributions that are capable of very sharp separations based on differences in the molecular kinetic diameter. Adsorbents made of activated alumina, activated carbon and silica gel separate by differences in adsorption equilibria, which must be determined experimentally. To be economical the adsorbent must be regenerable. This requirement excludes the processing of gas mixtures that contain: 1. High-boiling organic compounds because they are difficult to remove 2. Lower-boiling organic compounds that may polymerize on the adsorbent surface 3. Highly acidic or basic compounds that may react with the adsorbent surface

346

12 Separation method selection

Molecular sieves are excellent adsorbents for selective water removal. Solid-phase desiccant systems are relatively simple to design and operate. Generally they are the lowest cost alternative for processing small quantities of gas. For bulk separations, the more strongly adsorbed component should be in minority in the feed. Some large-scale enrichments examples include hydrogen recovery, methane enrichment and oxygen enrichment from air. In purification separations adsorption should only be considered for dilute solutions. Activated carbon adsorption is generally uneconomical for the removal of >1000 ppm contaminants. This is due to the fact that the costs of regeneration become too excessive in those cases. Regeneration can be performed by steam stripping or burning. During burning part of the active coal is lost, resulting in excessive cost when large amounts of contaminants have to be handled. 12.4.2.3 Membrane separation The separation factor defined earlier can be applied to gas permeation if the relative volatility is replaced by the permselectivity (see Eq. 1.11), which is the ratio of the membrane permeabilities for the two key components of the feed-gas mixture. This permeability depends on both solubility and diffusivity through the membrane at the conditions of temperature and pressure. Permeabilities and permselectivities are best determined by laboratory experiments. In general, gas permeation is commercially feasible when the permselectivity between the two components is greater than 15. Commercial applications include the recovery of carbon dioxide from hydrocarbons, the adjustment of the hydrogen-carbon monoxide ratio in synthesis gas, the recovery of hydrocarbons from hydrogen and the separation of air into nitrogen- and oxygen-enriched streams.

12.5 Separation methods for solid-liquid mixtures Solid-liquid separation equipment selection is dependent on such a huge number of different factors and the choice is so wide, that it is impossible to present a recipe that can be followed to select the right piece of equipment for each application. When developing a process many combinations of machine and technique are possible. Typical for solid-liquid separations is that several of these combinations will result in an adequate solution to the problem, while obtaining the optimal solution would inevitably too expensive, if not impossible in an industrial situation. A first step in the selection process is to choose the most appropriate technology from sedimentation, filtration or a combination of these two operations. The use of gravity sedimentation systems would generally be considered first because they tend to be relatively cheap, and are ideally suited for large continuous liquid flows and automatic operation. Settlers can also be used for classifying particles by size

12.5 Separation methods for solid-liquid mixtures

347

Table 12.5 Selection of sedimentation machinery.

Selection Criterion

Process scale (m3/hr) Settling rate (cm/s) Operation Objective: clarified liquid concentrated solids washed solids

Gravity Sedimentation

Sedimentation Centrifuges

Hydrocyclones

Tubular Bowl

Disc

Scroll

1 − >100 0.1 − >5 batch or continuous

1 − 10 100 100 1 − >100 0.5 0.1 – >0.5 batch continuous or continuous

x x x

x x

x x

x x

x x x

or density, which is usually not possible with filtration. However, settling does not give a complete separation. One product is a concentrated suspension and the other is a liquid, which may contain fine particles of suspended solids. In spite of this limitation it is important to note that even if sedimentation does not achieve exactly the separation required, it may still be a valuable first step in a process and may be followed by filtration. Pre-concentration of the solids will reduce the quantity of liquid to be filtered and the concentrated suspension can then be filtered with smaller equipment than would be needed to filter the original dilute suspension. However, too small a difference in density between the solid and fluid phases would eliminate sedimentation as a possibility, unless this density difference can be enhanced, or the gravity force field increased by centrifugal action. Having decided on sedimentation, the next stage is to consider the different equipment available and whether the separation is to be effected continuously or discontinuously. Various attempts have been made to provide a basis for the selection of sedimentation equipment. Table 12.5 contains some of this information, which points to the principal factors influencing such selections. Opposite to sedimentation, filtration provides an almost unlimited amount of design control since it relies on the filter material, a man-made septum chosen to retain exactly those particles that must be separated. Filtration systems, however, are far less suitable for continuous production and tend to be more expensive per volume treated than sedimentation systems. Therefore filtration is mostly applied when one wants to obtain the solids in the form of a cake, either as the desired end product or in removing undesired particles from liquid products. Although even more complicated, several attempts have been made to develop a methodology for filtration equipment selection. For example the cake build-up rate criterion relates the rate of filtration to equipment types, as shown in Table 12.6. However, it must be realized that all such generalizations are merely tentative suggestions on filtration equipment selection. Many types of filters are capable of similar performance and could be interchanged.

348

12 Separation method selection Equipment selection and rate of filtration.

Table 12.6

System characteristic

Cake thickness build-up rate

Equipment

Rapid separation

cm/s

Medium separation

cm/min

Slow separation

cm/hr

Clarification

negligible cake

Gravity pan screens Top-feed vacuum filters Filtering centrifuge Vibratory conical bowl Vacuum drum, disc, belt Solid-bowl centrifuge Pressure filters Disc and tubular centrifuge Cartridges Granular beds Filter aids, precoat filters

Fig. 12.11

Application range of various membrane filtration processes.

For the separation of very fine particles down to molecular size, membrane filtration techniques such as microfiltration, ultrafiltration, nanofiltration and reverse osmosis should be considered. The range of application of these four pressure driven membrane separation processes is summarized in Fig. 12.11. Ultrafiltration and microfiltration are basically similar in that the mode of separation is molecular sieving through increasingly fine pores. Microfiltration membranes filter colloidal particles and bacteria from 0.1 to 10 μm in diameter. Ultrafiltration membranes can be used to filter dissolved macromolecules, such as proteins, from solutions. The mechanism of separation by reverse osmosis membranes is quite different. In reverse osmosis membranes the membrane pores are so small that they are within the range of thermal motion of the polymer chains that form the

Exercises

349

membrane. Nanofiltration membranes fall into a transition region between pure reverse osmosis and pure ultrafiltration. Both techniques are used when low molecular weight solutes such as inorganic salts or small organic molecules have to be separated from a solvent. The most important application of reverse osmosis is the desalination of brackish groundwater or seawater.

Exercises 1 The vent gas from several polymer plants contains air, paraffins and olefins (monomers). The composition is given by (1 bar, 25 °C): Air Butane Pentane Hexane

91 1.5 1.0 0.5

mol% mol % mol % mol %

1-Butene 3 mol % 1-Pentene 2 mol % 1-Hexene 1 mol %

Within the R&D department of the company it is explored which techniques may be technically and economically feasible to recover the three olefins (1-butene, 1-pentene en 1-hexene) in pure state. The company has decided to approach you as a consultant and prepared several questions for the first meeting. As a first step they are considering the purification of the air by simultaneous recovery of the olefins and the paraffins. a. What separation methods are in your opinion applicable for this simultaneous olefin and paraffin recovery from the vent gas. After recovery the olefin/paraffin needs to be separated in two fractions, the paraffins (butane, pentane, hexane) and the olefins. This is in principle possible by selective absorption of the olefins. b. Can you determine which two key components need to be separated (selectivity > 1) in order to achieve a full olefin/paraffin separation. c. What is the minimal required difference in affinity (activity coefficients) of the absorption liquid towards both key components to achieve this full olefin/paraffin separation. In other words achieve a selectivity = (x1 / y1) / (x2 / y2) > 1. A possible alternative for absorption is extraction with a solvent that has a higher affinity for olefins compared to paraffins. d. What are in your eyes the two key components in order to obtain a full olefin/ paraffin separation when extraction with a highly polar solvent is applied? e. Which of the below solvents are in principle suitable (selectivity > 1) for a complete separation of the olefin and paraffin fractions.

Vapor Pressure (25 °C, bar) Distribution Coefficient (concentration in solvent/ concentration in feed) NMP DMSO DEG

C4

C4=

C5

C5=

C6

C6=

2.4

3.0

0.68

0.85

0.20

0.24

0.10 0.05 0.02

0.12 0.15 0.25

0.08 0.04 0.01

0.08 0.12 0.20

0.06 0.03 0.01

0.05 0.10 0.15

350

12 Separation method selection

2 In a large chemical plant methylcyclohexane is separated from toluene by distillation. For the design of a new plant both the use of extractive distillation and extraction are considered. A detailed evaluation requires first the characterization of the current separation between methylcyclohexane and toluene. a. Calculate from the given data the relative volatility for the separation of methylcyclohexane and toluene by distillation. Consider the mixture to behave almost ideal. b. Which of the given solvents might be feasible to enhance the relative volatility sufficiently to make extractive distillation economically attractive. c. The same question for extraction. Assume that the solvents and the methylcyclohexane/toluene mixture are almost immiscible. Besides methylcyclohexane (45 %) and toluene (45 %) the feed also contains 5 % of cyclohexane as well as benzene. d. What would be the optimal arrangement of separation steps for distillation, extractive distillation and extraction (for each 1) to separate this mixture into its four pure constituents? What distillative separation might become problematic and why. e. Which of the arrangements seems, considering the number of operations (columns), the most economical solution. f. Why is the counting of the number of operations sufficient to get a first impression about the most economical route? What effects are not taken into account? Benzene Vapor pressure (80 °C, bar) Activity coefficient at infinite dilution NMP DMSO DEG NMF

1.01

1.5 4.0 7.0 1.5

Cyclohexane 0.99

7 25 41 10

MethylCyclohexane 0.54

8 30 60 10

Toluene 0.39

2 4 10 2.5

Appendix Answers to exercises Please find detailed solutions at http://dx.doi.org/10.1515/9783110306729_suppl

1. Introduction 2.

α = 2.26

2. Evaporation and distillation 1.

A = 9.2082; B = 2755.6; C = 219.16 °C

2.

0 Pbenzene = 1.013bar

3 a. 100 °C and 1.0 atm: b. 100 °C and 1.5 atm:

x = 0.257, x = 0.736,

y = 0.456 y = 0.871

4. a. Tfeed = 369.9 K b. V/ L = 0.296 5. b. x = 0.318, c. x = 0.375, d. x1 = 0.36, x2 = 0.295, V1 = 0.4 F, 6. a. b. c. f. g.

y = L = y1 = y2 = V2 =

0.538 466.7 mol/h, V = 233.3 mol/h 0.585 0.511 0.18 F, L2 = 0.42 F

327.5 K V ″ = 101.4 mol/s 3042 kJ/s Nts = 11.5 including reboiler ≈6m

Nmin = 5.4 (Fenske) Rmin = 0.77 b. N = 16 + reboiler Feedstage = tray 4 from bottom

7. a.

8. a. b. c. d. e.

Rmin= 1.08 (Underwood) Nmin = 6.4 (Fenske) Nts = 11 + reboiler B = 487.3 kg/h D = 420 kg/h 255 kg/h

352

Appendix

3. Absorption and stripping 1.

xout = 6.44 × 10–6

2.

K=γ

P0 Ptot

= 0.682

For L / G = 1.5 * (L / G)min = 1.5 · 0.648 = 0.97, Nts ≈ 5.4 3.

(L / G)min = 80.5 at L / G = 120.8, Nts ≈ 4.4

4. a. Weight fraction base, (L / G)min = 0.471 b. Nts ≈ 8 5. a. K = 5 × 10–3 b. Use conventional mole fractions; N = 2.25 c. For L / G = 0.005, N = 8 For L / G = 0.004 the desired separation cannot be achieved because the minimum L / G ratio = 4.44 × 10–3 d. Nts (stripping) = 2.8 e. Nts (stripping) = 2.8; Nts (absorption) = 2.3 f. 0.1 < xin (absorber) < 0.2

4. G/L-Contractors 3. a. 2.5 m b. 0.14 c. 6.0 m 4. a. b. c. d. e.

Nts ≈ 3.7 EMV = 0.69 Ns = 4.4 xout = 0.16·10–6 Qmax = 3.4 m3 s–1; Qmin = 0.1 m3 s–1

5.

Fp = 540 m–1 Fp = 125 m–1

6. a. b. c. d. e.

20 % NOV = 3.62 kOV = 5.68 · 10–3 m s–1 uflood = 0.92 m s–1 20 % higher

uflood = 1.18 m s–1 uflood = 2.46 m s–1

5. Liquid-liquid extraction 1.

Graphical Nts = 4.1 Kremser Nts = 4.13

Answers to exercises

353

y1 = yout = 9.26 × 10–4 yN+1 = yin = 6 × 10–6 b. R′ = 51.6 (a.u.) xN′ = 0.0178

2. a.

3. a. b. c. d. e.

S/F S/F S/F S/F S/F

= = = = =

15.8 5.78 3.24 2.5 0.79

4. a. Xin = 0.13; Xout = 5.025 · 10–3; Yout = 0.105 c. S / F = 0.84 d. Nts = 3.9 5. a. Smin = 232 b. N = 4.5 c. Approximate solution:

6.

Extract

Ac

H2O

EA

kg/h

Raffinate Ac

H2O

EA

kg/h

E1 E2 E3 E4 E5

0.51 0.43 0.34 0.24 0.10

0.04 0.03 0.025 0.02 0.01

0.45 0.54 0.635 0.74 0.89

769 673 590 473 350

R1 R2 R3 R4 R5

0.59 0.66 0.74 0.84 0.945

0.03 0.02 0.02 0.01 0.005

914 809 714 622 500

Feed composition: 45 % benzene, 55 % TMA, 0 % H2O S / F = 0.56

7. a. Smin = 27.3 ton/h b. xC = 0.22 c. xC = 0.46 9. a. b. c. d.

0.38 0.32 0.24 0.15 0.05

Smin = 0.073 kg s–1 E1 = 0.17 kg s–1 Nts ≈ 3 xE2pyr = 0.10; E2 = 0.12 kg s–1

6. Adsorption and ion exchange 3.

14.8 kg/m3

4. a1. a2. a3. b1. 6. b. d.

W / V = 0.180 kg/m3 V = 1003 m3 6.15 · 106 s (71 days) 2.14 kg/m3 200 s 222 s

354

Appendix

7. Drying of solids 1.

0.34 g/s m2

2.

3.1 h

3.

0.17 g/s (h = 54.1 W/m2 K)

4.

psat ∞ psat

= 1.011

5.

T > 332.6 K

6. a. c. d. e.

785 kg 50.4 h 175 kg 3.7 h

8. Crystallization 1. a. 0.436 b. 0.718 2. a. maximum production rate 3182 kg/h b. maximum yield 0.795 3.

63.6 kg/h crystals/h

4.

195 kg/h

5. a. b. c. d.

0.142 0.187 0.343 0.443

6.

ΔE = 8.2 kJ/mol

7.

s = 0.006

kg kg kg kg

Na2SO4 Na2SO4 Na2SO4 Na2SO4

· · · ·

10 10 10 10

aq aq aq aq

per per per per

kg kg kg kg

free free free free

water water water water

fed fed fed fed

8. a. G = 0.097 μm/s b. B 0 = 3.7 × 104 no/s kg 9. a. 1.9 mm b. G = 0.082 mm/h 10.

τ = 9740 s Vcrystallizer = Vsolution + Vcrystals = 27.6 + 1.02 = 29 m3

9. Sedimentation 1.

Rep = 0.3 v∞ = 4.27 × 10–3 m/s

Answers to exercises

2.

A = 11.8 m2

3.

v∞ = 0.034 m/s A = 8.7 m2

4.

A = 2.9 m2 D = 1.9 m (cylindrical tank)

5.

12.6 μm (check Re = 0.055) L1 + L2 = 28.4 m

6.

vg∞ = 1.75 × 10–6 m/s vc∞ = 0.98 × 10–2 m/s

7.

49 Hz (98 π rad/s)

8.

Σprocess = 918 m2 Σmachine = 3580 m2 efficiency = 25 %

9.

Q = 4.47 × 10–4 m3/s

10. a. 9.05 · 10–5 m3 s–1 b. 0.15 μm (Re < 1) 11.

D = 0.75 m 4 hydrocyclones in parallel

10. Filtration 1.

α = 4.67 × 1012 m/kg, Rm = 2.45 × 1012 1/m

2.

α = 5.23 × 1011 m/kg

3.

α = 5.97 × 1011 m/kg, Rm = 9.04 × 1011 1/m

4.

α0 = 8.11 × 108 m/kg, n = 0.38

5.

α0 = 2.36 × 108 m/kg, n = 0.574

6.

filtrate production rate = 3.15 · 10–5 m s–1 dry solid production rate = 3.15 · 10–3 kg m–2 s–1 cake thickness = 0.90 mm

7.

9 m2

8.

0.627 kg s–1

9.

–

10.

0.66 kg s–1

355

356

11.

Appendix

3.77 h

12. a. 31.1 s b. 2.75 times faster

11. Membrane filtration 2.

8.75 10–8 kmol/m2 s 0.0075 and 0.045 kmol/m3

3.

7.45 10–8 kmol/m2 s 13 and 8.7 mol/m3

4.

4.88 atm

5.

PM,water / δM = 2.07 10–4 kg/m2.s.atm PM,salt / δM = 3.90 10–7 m/s

12. Separation method selection 1.

2.

a. Adsorption, absorption, condensation b. 1-butene and hexane γhexaan > 15 c. S > 1 ⇒ γ1−buteen d. 1-hexene and butane e. S > 1 NMP S = 0.5 (not suited) DMSO S = 2.0 (OK) DEG S = 7.5 (OK) P1* 0.54 a. α12 = * = = 1.38 P2 0.39 b. NMP α12 = 5.52 DMSO α12 = 10.4 DEG α12 = 8.28 NMF α12 = 5.52 For αD = 1.38 economics extractive vs normal distillation demand αED ≥ 2 Conclusion: all four solvents should be evaluated c. NMP β12 = 4.0 DMSO β12 = 7.5 DEG β12 = 6.0 NMF β12 = 4.0 For αD = 1.38 economics extraction vs distillation demand αE ≥ 5 Conclusion: only DMSO and DEG interesting for evaluation

Answers to exercises

357

d.

Benzene/Cycohexane separation α12 = 1.02, difficult, large chance for azeotrope. e.

Distillation

References and further reading

General 1. C. J. Geankoplis, Transport Processes and Separation Process Principles, 4th Edition, Prentice Hall, Englewood Cliffs NJ, 2003. 2. R. E. Treybal, Mass-Transfer Operations, 3rd Edition, McGraw-Hill, New York, 1980. 3. J. L. Humphrey and G. E. Keller II, Separation Process Technology, McGraw-Hill, New York, 1997. 4. F. M. Khoury, Predicting the Performace of Multistage Separation Processes, 2nd Edition, CRC Press, Boca Raton Fl, 2000. 5. C. J. King, Separation Processes, 2nd Edition, McGraw-Hill, New York, 1980. 6. W. L. McCabe, J. C. Smith and P. Harriot, Unit Operations of Chemical Engineering, 7th Edition, McGraw-Hill, New York, 2005. 7. R. D. Noble and P. A. Terry, Principles of Chemical Separations with Environmental Applications, Cambridge University Press, New York, 2004. 8. K. Sattler and H. J. Feindt, Thermal Separation Processes, Principles and Design, VCH, Weinheim, 1995. 9. S. J. Setford, A Basic Introduction to Separation Science, Rapra Technology, Shawbury, 1995. 10. E. J. Henley, J. D. Seader and D. K. Roper, Separation Process Principles, International Student Version, 3rd Edition, John Wiley & Sons, Hoboken, NJ, 2011. 11. P. C. Wankat, Rate-Controlled Separations, Springer, 1994. 12. P. C. Wankat, Separation Process Engineering: Includes Mass Transfer Analysis, 3rd Edition, Prentice Hall, Upper Saddle River NJ, 2012. 13. J. F. Richardson and J. H. Harker, with J. R. Backhurst, Coulson and Richardsonʼs Chemical Engineering, Volume 2, Particle Technology and Separation Processes, 5th Edition, Butterworth-Heinemann, Oxford, 2002. 14. R. K. Sinnot and G. Towler, Coulson and Richardsonʼs Chemical Engineering, Volume 6, Chemical Engineering Design, 5th Edition, Butterworth-Heinemann, Oxford, 2009. 15. Handbook of Separation Process Technology, (R. W. Rousseau, Editor), John Wiley & Sons, New York, 1987. 16. Perry’s Chemical Engineers Handbook, 8th Edition (R. H. Perry and D. W. Green, Editors), McGraw-Hill, New York, 2008. 17. Encyclopedia of Separation Technology, (D. M. Ruthven, Editor), John Wiley & Sons, New York, 1997. 18. Handbook of Separation Techniques for Chemical Engineers, 3rd Edition (P. A. Schweitzer, Editor), McGraw-Hill, New York, 1997. 19. Handbook of Chemical Engineering Calculations, 4th Edition (T. G. Hicks and N. P. Chopey, Editors), McGraw-Hill, New York, 2012. 20. Ullmann's Encyclopedia of Industrial Chemistry, 7th Edition, Wiley-Interscience, New York, 2002–. 21. Kirk-Othmer Encyclopedia of Chemical Technology, 5th Edition (A. Seidel, Editor), John Wiley & Sons, Hoboken, 2004–. 22. Kirk-Othmer Separation Technology, 2nd Edition (R. Clavier, Editor), Wiley, 2008.

360

References and further reading

Chapter 2 and 4 23. M. J. Lockett, Distillation Tray Fundamentals, Cambridge University Press, Cambridge, 2009. 24. H. Z. Kister, Distillation Operation, McGraw-Hill, New York, 1990. 25. H. Z. Kister, Distillation Design, McGraw-Hill, New York, 1992. 26. J. G. Stichlmair, J. R. Fair, Distillation: Principles and Practice, Wiley-VCH, New York, 1998. 27. R. F. Strigle, Packed Tower Design and Applications: Random and Structured Packings, 2nd Edition, Gulf Publishing Company, Houston, 1994. 28. J. Maćkowiak, Fluid Dynamics of Packed Columns, Springer, Heidelberg, 2010. Among many other reference data books, a lot of vapor-liquid data can be found in: 29. B. D. Smith and R. Srivastava, Thermodynamic Data for Pure Compounds, Elsevier, 1986. 30. B. E. Poling, J. M. Prausnitz and J. P. OʼConnell, The Properties of Gases and Liquid, 5th Edition, McGraw-Hill, New York, 2000. 31. C. L. Yaws, Handbook of Vapor Pressure: Antoine Coefficients, Gulf Publishing Co., Houston, 2007. 32. J. G. Speight, Lange’s Handbook of Chemistry, 70th Anniversary Edition, McGrawHill, New York, 2004. 33. J. Gmehling and U. Onken, Vapor-Liquid Equilibrium Data Collection, Chemistry Data Series, Dechema, Frankfurt a. M., 2003. 34. C. L. Yaws, The Yaws handbook of thermodynamic properties for hydrocarbons and chemicals, Gulf Publishing Co., Houston, 2007.

Chapter 3 35. R. Zarzycki and A. Chacuk, Absorption: Fundamentals & Applications, Pergamon Press, Oxford, 1993. 36. F. A. Zenz, Design of Gas Absorption Towers, in Handbook of Techniques for Chemical Engineers, 3rd Edition (P. A. Schweitzer, Editor), McGraw-Hill, New York, 1997.

Chapter 5 37. R. E. Treybal, Liquid Extraction, 2nd Edition, McGraw-Hill, 1963. 38. Handbook of Solvent Extraction (T. C. Lo, M. H. I. Baird and C. Hanson, Editors), John Wiley & Sons, New York, 1983 (reprinted 1991). 39. Science and Practice of Liquid-Liquid Extraction (J. D. Thornton, Editor), Volume 1&2, Oxford Science Publications, Oxford, 1992. 40. Liquid-Liquid Extraction Equipment (J. C. Godfrey and M. J. Slater, Editors), John Wiley & Sons, New York, 1994. 41. Solvent Extraction Principles and Practice, Revised and Expanded (J. Rydberg, M. Cox, C. Musikas and G. R. Choppin, Editors), 2nd Edition edited by J. Rydberg, CRC Press, 2004.

Chapter 6 42. D. M. Ruthven, Principles of Adsorption and Adsorption Processes, Wiley, New York, 1984. 43. D. Naden and M. Streat, Ion Exchange Technology, Ellis Horwood Series of the Society of the Chemical Industry, London, 1984.

References and further reading

361

44. A. E. Rodrigues, Ion Exchange: Science and Technology, NATO ASI Series E107, Martinus Nijhoff, Boston, 1986. 45. M. Suzuki, Adsorption Engineering, Kodansha Elsevier, Tokyo, 1990. 46. D. M. Ruthven, S. Farooq and K. S. Knaebel, Pressure Swing Adsorption, WileyVCH, New York, 1993. 47. C. E. Harland, Ion Exchange: Theory & Practice, 2nd Edition, RSC Publishing, 1994. 48. R. T. Yang, Gas Separation by Adsorption Processes, World Scientific Publ., Singapore, 1997. 49. B. Crittenden and W. J. Thomas, Adsorption Technology and Design, ButterworthHeineman, Oxford, 1998. 50. F. Rouquerol, J. Rouquerol and K. Sing, Adsorption by Powders & Porous Solids: Principles, Methods and Applications, Academic Press, San Diego, 1998.

Chapter 7 51. G. Nonhebel and A. A. H. Moss, Drying of Solids in the Chemical Industry, CRC Press, Cleveland, 1971. 52. R. B. Keey, Drying Principles and Practice, International Series of Monographs in Chemical Engineering, Volume 13, Pergamon Press, Oxford, 1972. 53. C. Strumillo and T. Kudra, Drying: Principles, Applications and Design, Topics in Chemical Engineering, Volume 3, Gordon & Breach Science Publishers, New York, 1986. 54. Handbook of Industrial Drying, 3rd Edition (A. S. Mujumdar, Editor), CRC Press, Boca Raton Fl, 2006. 55. Advances in Drying (A. S. Mujumdar, Editor), Hemisphere Publ. Co., Washington DC 1980–. 56. R. B. Keey, Introduction to Industrial Drying Operations, Pergamon Press, Oxford, 1978. 57. C. M. vanʼt Land, Industrial Drying Equipment, Marcel Dekker, New York, 1991

Chapter 8 58. N. S. Tavare, Industrial Crystallization, Process Simulation Analysis and Design, Plenum Press, New York, 1995. 59. Crystallization Processes, (H. Ohtaki Editor), Wiley, Chichester, 1997. 60. Crystallization Technology Handbook, 2nd Edition (A. Mersmann, Editor), Marcel Dekker, New York, 2001. 61. J. W. Mullin, Crystallization, 4th Edition, Butterworth Heinemann, New York, 2001. 62. Handbook of Industrial Crystallization, (A. S. Myerson, Editor), Butterworth-Heinemann, Boston, 2001.

Chapter 9 and 10 63. A. Bürkholz, Droplet Separation, VCH-Verlag, Weinheim, 1989. 64. P. N. Cheremisinoff, Solids/Liquids Separation, Technomic Publishing Co., Lancaster, 1995. 65. A. Rushton, A. S. Ward, R. G. Holdich, Solid-Liquid Filtration and Separation Technology, 2nd Edition, Wiley-VCH, 2000. 66. L. Svarovsky, Solid-Liquid Separation, 4th Edition, Butterworths, London, 2001. 67. W. Strauss, Industrial Gas Cleaning, Pergamon Press, Oxford, 1975.

362

References and further reading

68. D. B. Purchas, Solid/Liquid Separation Technology, Uplands Press, Croydon, 1981. 69. R. J. Wakeman, E. S. Tarleton, Filtration, Equipment Selection, Modelling and Process Simulation, Elsevier Advanced Technology, Oxford, 1999. 70. Handbook of Filter Media (D. B. Purchas and K. Sutherland, Editors), 2nd Edition, Elsevier Science, Oxford, 2002.

Chapter 11 71. J. L. Humphrey and G. E. Keller II, Separation Process Technology, Chapter 5, McGraw-Hill, New York, 1997. 72. M. Mulder, Basic Principles of Membrane Technology, 2nd Edition; Kluwer Academic Publishers, Dordrecht, 1996. 73. K. Scott, Handbook of Industrial Membranes, Elsevier Advanced Technology, Oxford, 1998. 74. Membrane Technology in the Chemical Industry, (S. Pereira Nunes and K.-V. Peinemann, Editors), Wiley-VCH, Weinheim, 2006. 75. R. Baker, Membrane Technology and Applications, 3rd Edition, Wiley, 2012.

Chapter 12 76. H. R. Null, Selection of a Separation Process, in Handbook of Separation Technology, (R. W. Rousseau, Editor), John Wiley & Sons, New York, 1987, Chapter 22. 77. S. D. Barnicki, J. J. Siirola, Separations Process Synthesis, in Kirk-Othmer Encyclopedia of Chemical Technology, 5th Edition (A. Seidel, Editor), John Wiley & Sons, Hoboken, 2004–. 78. W. D. Seider, J. D. Seader, D. R. Lewin, Product and Process Design Principles: Synthesis, Analysis and Evaluation, 2nd Edition, John Wiley & Sons, New York, 2004.

Index

absorbates 53 absorbent(s) 53, 57 absorber – classification 68 – design 56 – selection 71 absorption 53, 325, 327, 331, 343, 344, 345, 349 – factor 66, 103 – processes 55 acceleration – centrifugal 261 activated – alumina 147 – carbon 147 activity coefficients 19, 114 adiabatic cooling line 182 adiabatic-saturation temperature 181 adsorbates 143 adsorbent 12, 143, 145 adsorption 12, 143, 331, 345 – constants 13 – cycles 162 – isotherm 150 – selective 341 adsorptive capacity 145 aging 233 agitated contactors 131 air separation 166 ammonia production 55 amount adsorbed 158 analytical solution 65 anhydrous ethanol 47 anion resin 145 antibiotics 111 antisolvents 233 antoine equation 21 arrhenius – equation 152, 228 – relationship 58 axial dispersion 162 azeotrope 20, 46, 47 azeotropic distillation 46

backmixing 72, 130 balance envelope 29, 31 barrier 5 batch distillation 44 blowing 92 boiling point(s) 327, 337 – diagram 16 – overlapping 340 boilup 27 bound and unbound water 184 bowl centrifuge 254 break through point 156 breakthrough curve 158 Brinkman equation 247 Brownian motion 296 Brunauer classification 150 bubble – columns 70, 72 – point 16 bubbling area 93 cake – compressible 278 – – characteristics 278 – filtration 271, 275 – – constant pressure 280 – – dead-end 271 – incompressible 275 – pressure drop 278 – resistance 275, 278 – – specific 275, 278 calandria 216 capability 254 capacity 328, 334 capillary porosity 188 caprolactam 326 – synthesis of 326 carbon molecular sieve 155 Carman-Kozeny equation 248 carrier 112 cation resin 145 centrifugal – acceleration 251 – extractors 134

364

Index

– field 251 centrifuge 288 – bowl 255 – capacity of disc 258 – conical 288 – filtrate rate 291, 292 – fixed-bed 288 – multi-chamber 253 – peeler 288 – perforated basket 288 – pusher 289 – schematics 290 – scroll 253 – sedimentating 251 – sedimenting 252 – tubular 252 charge – saturation 264 chemical – absorption 53 – processes 1 Chilton-Colburn transfer numbers 181 choking 92 clarification 242 clarifier 250 – height of 250 clarifying disc stack 254 class I system 221 class II system 221 classification 8, 242 – of drying operations 192 classified – fines removal 218 – product removal 219 clear-liquor advance 218 coagulation 273 cocurrent cascade 118 co-current flow 319 coefficient of variation 211 collecting efficiency 262, 265, 266 collection efficiency 249, 256, 257, 262, 267, 294, 296 – single wire 294 column – diameter 77, 91 – height 77, 101, 103 – selection 107 combined heat and mass transfer 202 composite adsorbent 154 composition dependence 20 compressive front 159 concentrate see retentate

concentration – interfacial 227 – polarization 313, 314, 315, 316 – swing 162 – wave front 159 condenser 27 coning 92 constant – molar flow 30 – pattern 161 – separation factor 154 contact area 82 contaminants 336 continuous ion exchange 170 cooling crystallizers 214 countercurrent 53 – arrangement 120 – cascade 26, 121 – flow 130, 319 cross flow 319 crosscurrent cascade 119 crystal – characteristics 209 – formation 224 – growth – – mass transport 227 – – model 227 – – rate 210, 223, 226, 231 – production rate 221 – size 211 – – distribution 211, 218, 232 – – dominant 224, 232 crystallization 207, 339, 342 – advantages 209 – considerations in favour of 209 – design strategy 208 – disadvantages 209 – mass balance 220 – operating modes 212 – product requirements 209 – schematics 208 – staged or batch 214 – yield calculation 219 crystallizer 207, 342 – continuous 217 – classification 213 – design 207 – draft-tube agitated 217 – evaporating 215 – examples 216 – fluidized bed 217 – forced circulation 215, 217 – mixing in 214

Index

– modeling and design 221 – multiple-effect 216 – perfectly mixed 218, 230 – seeded batch 225 – vacuum 215, 216 CSD see crystal size distribution Daltonʼs law 17 Darcyʼs law 274, 275, 277, 290 DCC see direct-contact cooling decolorization 145 degradation 337 deionization 145 – processes 170 demineralization 145 demister 22 design reliability 329 desorption 54, 143, 167 – time 163 dew point 16 dialysis 301 difference point 126 differential distillation 21 diffusion 77 – charging 264 – equimolar counter 309 – in porous membranes 309 – Knudsen 310 diffusivities 79 – ideal gases 310 direct-contact cooling 215 – examples 215 dispersed phase 130 dispersive 163 – front 159 displacement purge 167 distillation 28, 11, 16, 77, 112, 331, 333, 334, 335, 336 – column 27 – conditions 29 – configurations 45 – costs 39 – sequence 48 distribution see crystal size distribution – coefficient 17, 58, 112, 114 downcomer 28, 70, 92 – flooding 92, 93 drag force 252 driving force 82, 178, 308 – and flux 315 dryer – batch 194

– belt 195 – contact 199 – direct-heat 194 – flash 196 – fluid-bed 200 – fluidized-bed 196 – rotary 195 – spray 198 – tray 194 – vacuum 199 drying 165, 331 – batch 192 – contact 193 – continuous 192 – convective 193 – dielectric 201 – freeze 201 – mechanisms 188 – radiation heating 201 – rate 178, 184, 188, 190 – – constant 189 – – curve 192 – – falling 190 – – mean logarithmic 192 – – proportionality 190 – steps 178 – sublimation 201 – time – – estimation 190 – – total 192 dumping 92 economics 327 effective diffusion coefficient 162 efficiency curve 249 electrodes – collecting 264 – collector 264, 265, 266 electrostatic precipitation 263 encrustation 215 energy – balance 43 – costs 335 – separating agent 5 enrichment – of air 346 entrainer 46 entrainment 92 – eliminator 23 – flooding 92, 95 equilibrium – based processes 4

365

366

Index

– charge 264 – controlled adsorbents 145 – line 63 – model 158 – phases 36 – ratio 11 – selectivity 153 – stage construction 127 – stages 22, 66 equimolecular diffusion 79 ergun equation 107 estimation of drying time 190 ethanol azeotrope 47 ethylene-oxide 1 eutectic 212 – phase diagram 213 evaporating crystallizers 215 evaporation rate 189 evaporative separations 15 evaporator 216 – multiple-effect 216 exploitable properties 7 external fields 5 extract 112 extraction 12, 326, 327, 332, 336, 340, 341, 349, 350 – factor 112, 118 – liquid-liquid 331 – scheme 117 extractive distillation 6, 46 extractors 130 extractor selection 135 feasibility 328 feasible operating region 104 feed 336 – composition 25 – line 35 – stage 33 Fenske equation 40 Fickʼs law 309 field – electric 264 – charging 264 film – mass transfer coefficient 227 – – theory 79 – theory 79 thickness 79 filter – aid 273, 287 – batch vacuum 283 – centrifugal 287

– continuous large-scal 279 – continuous vacuum 279 – deep-bed 293 – elements 285 – horizontal, belt 281, 282 – large-scale vacuum 279 – medium 286 – – resistance 274, 281 – Nutsche 283 – pressure 284, 285 – rotary disc 281 – rotating vacuum drum 280, 281 – screw 286 – vacuum 279 – variable-chamber 284 – wave plate 297 – wire mesh 295 – with tubular or flat filter elements 284 filtrate see permeate filtrate velocity 275 filtration 347 – centrifugal 272, 287 – constant pressure 277, 278 – constant rate 277, 279 – cross-flow 271 – deep bed 273, 287, 292, 293 – equipment 272, 279 – in centrifuges 289 – interceptive 292 – fundamentals 274 – micro 301 – of gases 293, 298 – process 271 – rate of 348 – ultra 301 – vacuum 272 fixed bed adsorption 156 flash – distillation 22 – drum 25 flashing 216 flocculant 250 flocculation 273 flooding 92, 104 – velocity 94 flow – parameter 92, 105 – rate 256 – sheets 1 flux – and driving force 315 – definitions 307 – equations 307

Index

force – acceleration 245 – buoyancy 245 – drag 245, 264 – electrostatic 264 – gravitational 245 fouling – column 337 fraction – absorbed 67 – extracted 118, 121, 124 – feed remaining liquid 24 – feed vaporized 24 – stripped 68 fractionation column 44 friction coefficient 94 functions 2 gas – number 92 – purification 55 – solubility 54 gas-liquid – contactors 77 – equilibria 57 g-factor 251 Gibbsʼ phase rule 23 grade efficiency curve 249 gravity-settling 242 growth rate – constant 228 – dispersion 226 – of crystal faces 210 – power-law 228 heat – and mass transfer – – combined 202 – capacity – – moist air 180 – duties 43 heavy key 41 height equivalent – of a theoretical plate 101 – to a theoretical stage 101 height of a transfer unit 103 Henryʼs law 12, 58, 152 heterogeneous feed 4 HETP 101 heuristic rules 45 Higgins loop contactor 170 high-boiling solvent 46 Himsley contactor 172

holdup of crystals 232 homogeneous – adsorbent 154 – mixtures 4 humidity 182 – absolute 180 – definitions 178 – relative 180 – saturation 181 – unit 179 hydrocyclones 259 – design 261 – in series 263 – schematics 260 ideal – absorbent 56 – liquid solutions 11 – non-equilibrium mixer 129 immiscible liquid 12 impingement see filtration of gases impurity removal 4 industrial – absorbers 68 – separations 10 inert-purge 167 inherent selectivity 11 interception – flow-line 296 – inertial 295 – mechanisms 295 interfacial – area 79 – concentration 227 – tension 224 inverse lever-arm rule 126 ion exchange 143, 167 – cycles 170 – resins 168 ion exchangers 145, 167 isolation 4 isomerisation 339 isotherm 12 isothermal flash 23 Karr column 132 Kelvin equation 187, 224 kinetic – selectivity 149, 155 – kinetic theory – – ideal gases 310 Knudsen – diffusion 155, 310 – flow 308

367

368

Index

Kremser equation 67, 89, 121, 124 Kuhni contactors 132 laboratory distillation 22 Langmuir – adsorption constant 153 – adsorption isotherm 151 – multicomponent model 153 Laplace equation 185 light key 41 linear – driving force model 161 – isotherm 152 liquid – diffusion 190 – distributors 69, 98 – feed fraction 34 – maldistribution 98 – redistributors 98 – residence time 72 liquid-liquid – equilibria 113 – extraction 111 loaded solvent 54 location of the feed stage 36 macropores 147 magma 216, 232 – concentration 226 mass – balance envelope 23 – concentrations 125 – discharge rate 219 – feed rate 219 – separating agent 5, 143 – transfer 6, 77, 129, 161 – – coefficient 79, 227, 314 – – zone 158 maturity 10 – of separation processes 329 maximum – allowable vapor flow rate 91 – permissible 249 – boiling azeotrope 21 McCabe-Thiele 121 – construction 38 – diagram 60 – method 31, 56 mean free path 310 mechanical separation(s) 8 – processes 4 melt – crystallization 233 – – countercurrent contacting 234

– crystallizer 235 – – examples 236 – – MWB 236 – – Sulzer MWB 235 membrane(s) – asymmetric 303, 304 – ceramic 303 – dense 310, 311 – filtration – – classification of 305 – – processes 304 – fouling 322 – modules 316 – – applicability 318 – – co-current 319 – – complications 319 – – counter current 319 – – cross flow 319 – – cut 321 – – design procedure 320 – – flow patterns see co-current flow, see counter current flow, see cross flow – – hollow fiber 317, 318 – – plate-and-frame 316 – – size and capacity 318 – – spiral wound 316, 317 – – tubular 316, 318 – porous 309 – selection 302 – separation 346 – – disadvantages 322 – symmetrical 303 – transport mechanisms 309 mesopores 147 metals processing 111 metastability 221 microfiltration 301, 304, 348 microporous 145 minimal required 248 minimum – absorbent flow rate 59 – boiling azeotrope 21 – liquid load 104 – number of theoretical stages 38, 40 – reflux 39 – – ratio 38 – solute free solvent flow 124 – solvent flow rate 123 – solvent-to-feed ratio 128 – stripping gas flow rate 60 mixed bed systems 170 mixed-suspension mixed-product removal crystallizer 229

Index

mixer-settler 130 moisture – content 180, 184, 187, 188, 190, 191, 202, 203, 204 – – critical 190, 191 – free 189 – in solids 183 – kinds of 184 molar heat of evaporation 34 mole ratios 61 molecular – diffusion 79, 155 – properties 7 – separation(s) 9 – – methods 10 – – processes 4 – sieve 147 – sieving 149 – weight cut-off 312 morphology 209, 210 MSMPR see mixed-suspension mixedproduct removal 229 multi-chamber 253 multicomponent mixtures 25, 45 multiple liquid mixtures 45 Murphree – efficiency 43, 87 – vapor efficiency 43 MWB melt crystallizer 236 MWCO, see molecular weight cut-off nanofiltration 301, 303, 305, 348 naphtha crack 325 Newtonʼs second law 245 nitric acid 55 non-equilibrium trays 89 non-ideal 19 nonsolvents 233 nucleation 223 – contact 225 – homogeneous 223 – primary 223 – rate 223, 224, 225, 231, 232 – – secondary 225 – secondary 223, 225 nucleus 207 number density 229, 231 number of – equilibrium stages 36, 64 – stages 125 – theoretical stages 29, 77, 124 – transfer units 103 Nusselt number 181

oil refining 1 Oldshue-Rushton column 132 operating – capacity 169 – lines 24, 31, 32, 61, 121, 126 – point 126 – pressure 29 – temperatures 29 – window 72 optimal – feed stage location 37 – operating reflux ratios 39 – separation sequence 45 OSLO crystallizer 217, 218 osmotic pressure 306, 316, 322 overall – composition 125 – efficiency 86, 89 – flow rates 61 – mass balance 29 – mass-transfer coefficient 81 – number of gas-phase transfer units 88 – rate constant 228 oxygen nitrogen separation 156 packed – bed – – flow through 274 – columns 69, 72, 97, 133 packing 69, 97 – factor 105 pall ring 99 partial – evaporation 15 – pressure 18 partially miscible systems 124 penetrant see permeate perforated-plate columns 134 permeabilities 13 permeability 308, 309, 310, 346 permeance 308 permeate 302 permeation 13 – vapour 301 permselectivity 346 pervaporation 301 physical absorption 53 plait point 115 plate – columns 70, 83 – distillation column 27 plate-and-frame 284

369

370

Index

Podbielniak extractor 135 Poiseuille flow 155 polymorphism 210 population – balance 208, 229, 230, 231 – density 229, 230, 231, 236, 236, 236, 238 – distribution 211 pore size distribution 155 porosity 274 potable water 307 Prandtl number 181 precipitation 209, 232 – examples 233 precipitator 264 – electrostatic 265, 267 – – single-stage 266 – single stage 265 – tube-type 266 – two-stage 265 press see filter pressure – drop 72, 96, 104, 106 – sensitive azeotropes 47 – swing 162 – – adsorpton 165 – – distillation 47 primary nucleation – heterogeneous 223 – homogeneous 223 process simulation 9 product – recovery 55 – stability 329 production rate – dry solids 281, 283 – filtrate 281, 283 productivity 279 propagation velocity 159 property differences 7 psychrometric chart 182 – of air-water 183 pulsation device 134 pulsed-packed columns 134 pulsed-plate column 134 purge swing adsorption 166 purification 4 q-line 35 raffinate 112 random packing 70, 98 Raoultʼs law 11, 18, 58

Rashig ring 98 rate constant 228 – crystal growth 228 rate law – parabolic 277 rate-controlled processes 4 real number of stages 90 reboiler 27, 28 reciprocating-plate columns 132 recovery 4 – hydrogen 346 rectification section 28 rectifying section 31 recycling 4 reflux 27 – ratio 28, 31 regeneration 143, 168 relative – humidity 182 – volatility 12, 17 required steam rate 43 residence time 231, 248, 255, 256, 259 – distribution 218 residue see retentate retentate 302 retention 312, 315 reverse osmosis 301, 303, 305, 307, 316, 348 Reynolds number 245, 246, 247, 315 Richardson and Zaki equation 247 rotating-disk contactor 131 saturation – solubility 212 – temperature adiabetic 182 – vapor pressure – – at curved interfaces 186 – – in pore 186 scale 334 scale-up 331, 335 scaling up membrane units 322 Scheibel column 131 Schmidt number 181, 315 screening 330 sedimentation 241, 347 – between inclined plates 249 – centrifugal 251 – clarification 241 – classification of equipment 241 – continuous 248 – critical point 244

Index

– dilute 243, 245 – effect of concentration 244 – graviity equipment 250 – gravity 242 – ideal continuous tank model 248 – in horizontal channels 249 – mechanisms 242 – schematics 243 – thickening 241 – time 255 seed crystals 225 selection 8, 325, 326, 327, 328, 329, 330, 331, 332, 333, 335, 342, 346, 347, 348 – scheme 135 selective retardation 303 selectivity 7, 114, 170 – and retention 312 – in dense membranes 312 – in gas separation 312 – in porous membranes 310 separating agent(s) 4, 56, 328 separation – bulk 346 – efficiency in hydrocyclones 262 – factors 6, 7, 11, 332, 346 – gases 301, 344 – in a disc stack 258 – liquid-liquid 241 – number 296 – of gases 303 – operations 2 – processes 1, 5 – solid liquid 241, 259 – techniques 1 – technology 1 separator – electrostatic single-stage 267 – lamellar plate 297 – plate-type 267 settler criterion 260 settling – gravity 248 – hindered 243, 247 – tank – – batch 250 – – continous 250 – velocity 246, 247, 249, 256, 261 – – terminal 248 – zone 217 Sheer breeding 225 Sherwood number 181, 315

shock wave 160 sieve tray(s) 28, 334 – extractor 133 sigma – concept 256, 262 – – scale-up 258 – parameter 256, 259 silica gel 147 single equilibrium stage 117 size – distribution 209 – exclusion 303, 308 slit 93 solid-liquid mixtures 346 solubility 213, 312 solute 112 – free 61 solution concentration – units 222 solution-diffusion 303, 308, 310, 311 solvent 112 – cost 116 – extraction 111 – recoverability 116 – recovery 112 – selection 115 – – criteria 115 sorption 143 specific – area 145 – surface area 146 spray – columns 132 – contactors 72 – towers 70 stage efficiency 129 staircase 36 steam chest see calandria still pot 22 stoichiometric – front 156 – time 156 Stokes 256 Stokesʼ law 246, 252, 260, 264 stripper 64 stripping 54, 64, 341 – factor 67 – section 28, 32 strong acid cation exchange resin 168 strong base anion exchange resin 168

371

372

Index

structured packing 70, 99 styrene-divinylbenzene copolymers 149 sugar 2 sulfonic acid resin 168 sulfuric acid 55 Sulzer MWB melt crystallizer 236 superficial – liquid velocity 92 – vapor velocity 92 – velocity 159 supersaturation 212, 214, 221, 222, 223, 224, 233 – and metastability 221 surface – diffusion 155 – tension 95 system air-water 180 technological asymptote 10 temperature-swing 162 terminal settling 246 theoretical trays 89 thermal – stability 29 – swing adsorption 164 thermodynamic equilibrium 6 thermosyphon 28 thickener 250 thickening 242 throughput – cyclone 262 – factors influencing 249 tie lines 114, 125 total reflux 38, 40 trajectory – critical particle 255 transfer units 87, 102 tray – columns 72, 83 – diameter 96 – efficiency 86 – sieve 28, 334 – spacing 86 triangular diagrams 113, 124 ultrafiltration 301, 303, 304, 348 unagitated columns 132 Underwood expression 41 urea 55 use asymptote 10

vacuum – crystallizers 215 – distillation 99 – filter 280 – – rotating drum 280 Vanʼt Hoff – equation 212 – relation 306 vapor – pressure 18 – liquid equilibrium 15 velocity – cross-stream drift 264 – migration 264, 265 – particle 251 – – drift 264 – radial 255 – tangential 261, 262 – terminal settling 247 void fraction 247 volatilities 15 volatility – low relative 336, 339, 337 – relative 334, 336, 337, 340 volumetric distribution coefficient 81 vortex – primary 259 – secondary 260 wave front 163 weak acid cation exchanger resin 168 weak base anion exchanger resin 169 Weber number 95 weeping 92 weight ratio 121 wet-bulb – cooling line 182 – temperature 179, 181, 183, 189, 202 wetting 104 xylene 339 x – y diagram 24 yield 219 – maximum 220 Yield Calculation 219 y – x diagram 16 zeolites 147 Zone melting 234